On 2018-07-19 00:26, Rob Lanphier wrote:

This sounds like a combination of majoritarian and minmax Approval. I
think that a two-round system should contain the winner of the first
round so that if the first round is correct, the winner is preserved.

So how about making that more explicit? Say something to the effect of:

The Approval winner advances, and
two other winners advance so as to maximize the number of ballots that
approve of at least one of them.

By having three instead of two, a center candidate won't bias the runoff
towards either the left or right in an LCR situation. On the other hand,
it's harder to justify once opinion space becomes multidimensional; and
having three instead of two does lead to the Approval dilemma of whether
to approve your favorite only or also approve your second best.

You could run the second round with a ranked method, but that would make
it a lot more complex. Letting two candidates pass instead of three
would solve the problem, as you could use Plurality for the second
stage, but any LCR situation would be biased: either it would be
left-leaning (LC), right-leaning (RC), or miss the centrist (LR).

From a design perspective, the best would probably be to have the
Approval winner advance, and as many additional candidates as are needed
to achieve proportional representation up to some set level (which would
act as an effective threshold). That suggests using a house-monotone PR
method to pick the candidates for the second round; but that would be
anything but simple.

If I recall correctly, choosing the candidates to maximize the number of
voters who approve of at least one of them is NP-hard for general n. See
https://en.wikipedia.org/wiki/Maximum_coverage_problem.

The best you can do without P = NP is a very obvious greedy algorithm
(successively pick the candidate that maximizes the number of additional
voters covered by the set of candidates so far). The greedy algorithm
makes it relatively easy to start with one or more candidates picked by
other means (e.g. the Approval winner) and then filling in the rest.

Thanks for the quick response!  I'll offer a quick response now, and
maybe a more detailed response later.

viable choice for California in the near-ish term, my bias is toward a
system that picks exactly two candidates as a replacement for our
current top two.  Using your "LCR" notation, I'm mildly biased toward
a system that errs on the side of selecting "LR" rather than "LC",
since I think that a lot voters are more likely to find "LR" to be
more fair than "LC", even though there will be some cases where "C"
would be the winner of a one-round approval contest.  My fear is that
with "LC", the supporters of "R" are more likely to declare the system
unfair, and push for a repeal.

That said, you're right that it's best to allow three candidate to
advance in the (hopefully) rare case that the two candidate set
doesn't already include the approval winner.  If this proposal is done
as a replacement for the current top-two system, that just means we
have a three-way plurality race.  I haven't decided if it'd be viable
to call for a change to approval voting for the general election as
well.  I'll save that for a future response (and please weigh in with
your thoughts)

I'll send a more detailed response later, I think...

Rob

On 2018-07-19 07:20, Rob Lanphier wrote:

I'd imagine that the more polarized things are, the more important it
would be to keep the centrist in the running, so that there's at least a
possibility of getting back into a less polarized state of things. E.g.
suppose L is quite far on the left and R is quite far on the right. But
you're right; it's impossible to do that with only two candidates if L
and R are strong enough, since any inclusion of the centrist would be
biased towards whatever wing got included.

If the second round method is plurality, and it's three-way, then the
centrist would probably lose in any case.

On Wed, Jul 18, 2018 at 4:37 PM, Kristofer Munsterhjelm <
km_elmet@t-online.de> wrote:

I believe runtime is exponential in the number of winners, so "2 advance"
and "3 advance" are quite tractable, and "5 advance" should be possible in
most cases.

Andy and Kristofer, thanks both of you for weighing in on this.
Kristofer: thanks for teaching me about the "maximum coverage
problem".  It's been a lonnnng time since I've been in a C.S./math
class to study that sort of thing.  But since I'm an unrepentant
packrat, I was able to find my textbook from when I was in C.S.
learning about similar problems (e.g. the four color theorem).  The
Wikipedia article was way more helpful  :-D  Andy, I suspect your
hunch is correct; if "2 advance" is all we compute, that doesn't seem
too difficult to achieve via brute force algorithm.

As I've been mulling this over and talking with other folks, I have a
hunch based on a number of assumptions:

1. The top-two primary system is ripe for reform
2. Change-averse voters want typical two candidate elections to
   include 1 Democrat, and 1 Republican, since that will seem most fair
   to them.  Change-averse Democrats don't mind when only 2 Democrats
   advance to the general, and change-averse Republicans don't mind when
   only two Republicans advance to the general, but the unfairness of the
   current system outrages them when the opposite party prevails.
3. Many of those change-averse voters dislike the top-two primary,
   for reasons described here:
   http://www.laweekly.com/news/states-top-two-primary-system-could-create-unintended-consequences-on-june-5-9511189
4. Many voters prefer having two elections (a primary and a general)
   and dislike efforts to conflate public consideration of candidates
   into a single all-purpose election
5. Many voters also don't mind having a diversity of choices in the
   primary, and dislike onerous restrictions on ballot access
6. There are many voters who abstain from voting in the primary
   election, but vote in the general.  I'll refer to them as
   "general-only voters",  General-only voters appreciate having a
   simplified choice in the general election, and are frustrated by
   having too many non-viable options that haven't been culled out.

My hunch: a ballot initiative focused on primary elections only is
more likely to be successful than one that attempts to modify the
general election.  As I've been mulling the large 22-week gap between
the primary and the general election here in California (in 2018, the
primary was Tuesday, June 5, general election is Tuesday, November 6),
I think the following could be a viable an incremental reform:
A. All candidates who receive over 50% approval qualify for the general election
B. The two candidates approved by the most number of voters also qualify
C. After the primary election, all qualified political parties have 28
days to endorse their preferred candidate.  Said endorsement is
required in order for a candidate to list the political party next to
their name on the ballot
D. After the primary election, all qualified candidates have 7 days to
accept/reject any official party endorsements.  Any accepted party
endorsements will be listed next to their name on the ballot.
E. Also 35 days after the primary election, qualified candidates may
withdraw their name from consideration in the general election.  As a
result of withdrawing, their name will not be printed as a candidate
in the general election.

Once again, some intentional ambiguity.  In the scenario I laid out in
my July 18 message: A-left, B-center, and C-right would all advance
(what Kristofer referred to as an "LCR situation").  Assuming the
Democratic/Republican duopoly still held, B-center would then be in a
good position to broker for the endorsement of either the Democrats or
the Republicans.  If none of the candidates drop out, then it's a
classic 3-way plurality race.

I think this LCR situation would be rare.  I believe a much more
common case would be that only one candidate gets over 50%, and that
that candidate is also part of the pair who qualifies in step "B"
above.  Thus it would be rare for three candidates to advance when
only one achieves 50% approval.  Moreover, in this system, all
candidates would be motivated to achieve >50% as their top priority to
guarantee a qualified spot on the general election ballot.

However, I wouldn't want to ignore the general election drama problems
that this system might create.  If this were proposed via ballot
initiative, I think some non-binding latitude+guidance that should
explicitly given to the state legislature should be:
a.  the ability to alter the general election, such that it also is
tallied via approval voting
b.  the ability to amend this, making the approval threshold lower for
advancing to the general election (e.g. "40%" or "30%" rather than
"50%")
c.  the ability to amend this, making it so that if only one candidate
gets 50% approval, they automatically win the general election (thus
making it a single-step approval election)
d.  the ability to amend this, making it so that the candidate
receiving the highest approval score wins the election, thus
consolidating the primary and general elections into a single
election.

All clauses above would make the glide path toward approval voting in
the general election clear, while making a change that even
change-averse voters can get behind (since the core of the initiative
would be a change to the top-two primary).  If my hunch #4 above is
correct, then some combination of clause "a" and/or clause "b" will be
the more popular alternative.  If I'm wrong, and most people like the
idea of having a single election to replace the primary+general system
we currently have, then clause "c" or even "d" will be the most
popular.

Does this proposal make sense, or is it too complicated to be viable?

Rob

On Thu, Jul 19, 2018 at 7:10 AM Andy Jennings
elections@jenningsstory.com wrote:

On 2018-07-24 00:03, Rob Lanphier wrote:

It might make the behavior seem too opaque to the voters if it just says
"try every possible alternate choice". On the other hand, voters'
intuition of how methods behave may not correspond to how they behave to
begin with; consider e.g. the chaotic nature of IRV.

I also happened across the following draft paper by James Green-Armytage
on the question of how to select candidates for a top-two runoff:
http://jamesgreenarmytage.com/runoff.pdf

It seems the best approaches he found use proportional representation
methods. For {L,R} in the LCR scenario, just running a 2-seat
proportional representation method; and for {L, C}, choosing an initial
winner by a majoritarian method, then use a 2-winner proportional
representation method to find the most representative set of two winners
subject to that one of them must be what the majoritarian method chose.

So the "deluxe" option might be: use a majoritarian method to find a
winner. Then use a proportional representation method to find a 2-seat
winning set. If the majoritarian winner is on that winning set, you're
done; use the set. Otherwise use the 2-seat set plus the majoritarian
winner (the classical three-way race you mentioned).

Picking the best PR 3-seat set subject to that one of the candidates in
it must be the majoritarian winner might be better, but on the other
hand, it might encourage tactical nomination. I'm unsure. Under honesty,
it would be somewhat better (e.g. if the majoritarian winner is slightly
to the left, it would deweight left-wing candidates more), but it might
not be worth the complexity.

There might even be a way to accommodate rules for when there's no need
to run a second round, if the PR method can quantify how well the
electorate is represented by the set. The idea would be that if the best
1-seat set (the Approval winner) already represents the electorate very
well, then there's no need for a runoff.

On Sat, Aug 11, 2018 at 1:50 PM Kristofer Munsterhjelm
km_elmet@t-online.de wrote:

Thanks for reviving this thread, Kristofer!  Also, thanks for pointing
out the Green-Armytage/Tideman paper.  I'll queue that up on my
reading list, and I'll be in a better position to respond to your full
mail.  Just in my initial skim, it looks like they make some of the
points that match my intuition on the subject.  In particular (to
quote the paper):

This is the reason why I dislike arguments that discount the value of
primary elections.  If structured correctly, a primary election
followed by a general election allows us to balance two competing
needs:
a.  we want elections with liberal ballot access as the most
democratic way to ensure a diverse selection of candidates.
b.  we don't want to overwhelm the electorate (and the press, and our
public debates) with too many candidates

A well-structured primary/general pairing can allow us to have both.
Primary voters can vet a wide field of candidates, and give us a sane
general election cycle.  With a more sophisticated voting system than
FPTP, it seems possible to have a field with more than 2 candidates
(e.g. 3, 4 or maybe even 5 candidates).  More than 5 candidates
becomes kind of a clown show, as evidenced by the 2008-2016 Republican
on-stage U.S. presidential debates, and the 2004-2008 Democratic
counterparts.  When the field narrows to 2-3 viable candidates, the
policy discussions can get interesting

I'd be skeptical of a system that declares no need for a second round,
and moreover, I suspect it won't be a viable incremental improvement
to California's existing system.  One path I could see forward is:

phase 0 (current): plurality-based top-two primary, plurality general
election with exactly 2 candidates
phase 1: approval-based primary, plurality general election which
almost always has 2 candidates
phase 1.1: approval-based primary, plurality general election which
frequently has 3-5 candidates
phase 2: approval-based primary, approval-based general election

The only difference between "phase 1" and "phase 1.1" is the
acclimation of the electorate and the politicians to the system,
rather than rule changes.  I imagine that, at first, voters would
bullet vote more frequently than strategically sensible, and
candidates would focus on traditional Democratic or Republican base
voters.  Over time, it seems as though candidates would figure out how
to exceed the 50% threshold as crossover candidates, and it would
become more commonplace (only dampened by fear of spoiler candidates
in the general election).  Phase 2 would require another change of
law, which my July 23 proposal tries to anticipate but not mandate.

Having experienced California elections up-close for a little while
(e.g. our 2018 U.S. Senate primary, which had 32 candidates), and
witnessing from a short distance the craziness of the 2003 California
gubernatorial recall election (which had 135 candidates), I've come to
appreciate the value of winnowing down the number of options.  I can't
imagine using a ranking system for that many candidates.  It will be
interesting to tear into the Green-Armytage/Tideman paper, but it may
require more brain cells than I have to spare tonight.  Soon, I hope
:-)

Rob

---------------------------- Original Message ----------------------------

Subject: Re: [EM] A simpler approval based way of replacing the CA jungle primary

From: "Rob Lanphier" robla@robla.net

Date: Tue, August 14, 2018 12:36 am

To: "Kristofer Munsterhjelm" km_elmet@t-online.de

Cc: "Election Methods" election-methods@lists.electorama.com

personally, i really dislike the California primary system that can potentially lock out a party from the general.� in fact, even though i dislike two-party hegemony as we have in the U.S., i actually believe that parties in politics can be and are a good thing (this doesn't mean
that there aren't bad parties or bad leadership in any particular party).� parties serve a useful purpose and, in my opinion, ballot access in the general election should depend solely on getting enough signatures and the state should recognize parties only to the extent of contract law to keep
some faction in a party from undemocratically screwing the rest of their party.� but every party should be able to get candidates on the General Election ballot, providing they get enough valid signatures on the ballot petition.��i don't think the General should be just between two
candidates.
and both the Primary and General Election should be decided with Ranked-Choice Voting using a Condorcet-compliant method (either Ranked-Pairs or Schulze) for the single-winner races.� STV or IRV is still sucky.
for Ranked-Choice voting, the number of candidates appearing
on the ballot should not greatly exceed the number of ranking levels.� when San Francisco has 20 candidates and 3 ranking levels, someone who voted their heart's select which included none of the real contenders will feel disenfranchised when they found out that the contender they hated the
most wins and they didn't vote for the second-place finisher to oppose that hated winner.� but the way to make this happen is not to put in 20 ranking levels, what is needed is sufficiently strict ballot access law that makes it harder to get on the ballot.� but not so hard that only one
or two can meet that requirement.
and, i think that the law should allow a candidate that loses in a primary or caucus to have some time after the primary to gather signatures to run as an independent candidate.
what i am still unsure of is what New York does.� New York allows a
single candidate to be nominated by more than one party and have that candidate's name appear multiple places on the ballot (associated with each party that nominated him/her) and then to total the votes for that candidate from all of the parties that he/she appears with on the ballot.�
George Pataki did this with the GOP and the Conservative Party nominations.� it can't be fair to the other candidates when one of them appears on the ballot more than one place.
ballot access is weird.� in Vermont, because lawmakers didn't anticipate this problem, a 14-year-old kid
is running for governor and appears on the Dem party primary ballot.� ya gotta be at least 18 to vote, but you need not be that to run.� ain't that weird?

--

r b-j� � � � � � � � � � � � �rbj@audioimagination.com

"Imagination is more important than knowledge."

�
�
�
�

Hi Robert,

Lots to think about here.  More inline:

On Mon, Aug 13, 2018 at 11:39 PM robert bristow-johnson
rbj@audioimagination.com wrote:

I think, up to this point, we largely agree.  I also dislike locking
out legitimate political parties, and I agree that political parties
provide useful mental shorthand for voters (even for wonks like
myself), but I'm willing to accept the risk of locking out candidates
and/or parties to achieve some of the goals from my earlier mail.
But, I agree, it's not good to lock out political parties (especially
important ones) from a general election.

Hmmm...I think California made a respectable tradeoff.  Maybe not the
best solution, but I (for one) appreciate the simpler ballot.
Moreover, I suspect that reform of the primary election is going to be
way more viable than trying to make a change to the general election
procedures.  If the government-funded primary remains an open
competition between all candidates (not just an intra-party affair)
with an improved voting method (like approval, or range, or even a
system that is explicitly-designed to select the Condorcet-winner as
one of the two candidates) and the general election is a final runoff
between two finalists, isn't that an acceptable compromise?  Or is it
a non-negotiable belief that all parties should get a candidate in the
general election?

I've been making a point of trotting out my Perl Journal article from
1996 (https://robla.net/1996/TPJ) at times like these, so as to
credibly confirm my shared desire for a Condorcet-compliant system.  I
think that after 22 years or so of an uncompromising push for
Condorcet-compliance, I'm about ready to compromise.  There are many
that argue that a well-implemented Approval system will pick the
Condorcet winner the vast majority of elections.  That seems credible
to me.

We agree, though:  IRV is sucky.  Standard STV seems inferior to
CPO-STV (though for filling a large-enough pool of seats, it seems
ok).

Yup.  I was happy with the 2018 outcome, but I talked to at least one
activist who abhorred the idea of selecting any of the three leading
candidates.  I think one of the big problems with San Francisco's RCV
is that this would have been considered an invalid ballot:

[1] [2] [3]
[ ] [x] [ ] Fringe candidate FCA
[ ] [x] [ ] Fringe candidate FCB
[ ] [ ] [ ] Mainstream candidate MCA
[ ] [x] [ ] Fringe candidate FCC
[ ] [ ] [x]  Mainstream candidate MCB
[ ] [ ] [x]  Mainstream candidate MCC
[x] [ ] [ ]  Fringe candidate FCD
[ ] [ ] [ ]  Fringe candidate FCE

i.e. prefers FCD, finds FCA, FCB, and FCC acceptable, and will
tolerate MCB or MCC, and believes that MCA is worse than just about
everyone else on the ballot (including most of the fringe candidates).
However, voters were only allowed to mark exactly one candidate per
column.  Just about all Condorcet-compliant ranked tally options don't
have a problem with tie ballots, and of course, score voting has no
problem with ties, but IRV can't deal with this.

California has done a lot to make ballot access stricter.  That's just
made it so that:

1. Working in downtown San Francisco is awful, because there is
   always a paid signature gatherer between the train stop and work
   (year-round, every year)
2. Frequently, the only candidates/initiatives that make it to the
   ballot are the ones with big money backing them.
3. There are rules that make it easy for an official to declare a
   ballot petition invalid.

I personally didn't mind that incumbent Dianne Feinstein had 31
opponents running against her for her U.S. Senate seat.  I just wish
we had a better method for picking the two candidates that would went
on to the general election.

Getting rid of FPTP in the general election is insanely hard.  Worse,
the IRV folks have way more momentum than people who like better
systems.  There's only one way that FPTP is an acceptable system, and
that is if there are only two candidates.  My proposals don't
guarantee that there are only two, but I think they make it rare
enough to be an acceptable risk.

Ditto.  Whatever it is, it's weird.  :-)

heh....that and this was apparently the first election that his dad
voted in  :-/
https://vtdigger.org/2018/08/15/hallquist-celebrates-historic-victory-trains-sights-scott/

Rob
p.s. Kristofer, I read the Green-Armytage/Tideman paper; more on that later...