[EM] Party-based top two with approval

[Rob Lanphier]

Hi folks,

The top two jungle primary system we use in California has a lot of folks
on edge. The worry among Democrats and Republicans is that, in some
districts, Democrats fear that we'll split the vote and end up with two
Republicans. Republicans fear splitting the vote and ending up with two
Democrats in the general election.

One possible alternative that seems more robust to me: for each office,
have a ballot as follows:

For Representative from [District X]:

Your preferred party:
[ ] Republican
[ ] Democratic
[ ] Libertarian
[ ] Green
[ ] Socialist

Candidate(s) you approve of:
[ ] Aaaaa (Republican)
[ ] Bbbbb (Republican)
[ ] Ccccc (Democrat)
[ ] Ddddd (Democrat)
[ ] Eeeee (Socialist)
[ ] Ffffff (Libertarian)

The candidate election would be approval based. The single candidate with
the highest approval rating (regardless of party affiliation) would be
guaranteed to make it to the general election. The ballots would also be
divided by party preference, such that each party's preferred candidate
would be declared the party nominee. The party with the most votes would
be able to choose whether they want to advance their nominee to the general
election. If they opt out (or if their top choice is the same as the
overall approval winner), then the next party gets to choose, until a
second candidate is chosen.

As an example, let's say we have 100 people voting with the ballot above.
We end up with this:

Aaaaa 45 votes (by 30 Republicans and 5 Libertarians and 10 Democrats)
Bbbbb 20 votes (by 15 Republicans and 5 Libertarians)
Ccccc 60 votes (by 35 Democrats, 15 Republicans, 5 Socialist, 5 Green)
Ddddd 50 votes (by 30 Democrats, 10 Socialists, 10 Green)
Eeeee 45 votes (by 10 Socialists, 30 Democrats, 5 Green)
Ffffff 30 votes (by 15 Libertarians, 15 Republicans)

The party split:
30 Republican
35 Democratic
15 Libertarian
10 Green
10 Socialist

Ccccc would advance as the general winner. The Democrats have the most
votes, so they have the option of advancing their top vote getter.
Aaaaa 20
Bbbbb 0
Ccccc 35
Ddddd 30
Eeeee 30
Ffffff 0

The self-identified Democrats also pick Ccccc, so we move to the next party.

There are 30 voters who self-identified as Republicans for this election.
Their approved nominee:
Aaaaa 30
Bbbbb 15
Ccccc 15
Ddddd 0
Eeeee 0
Ffffff 15

Among those 30 voters who self-identify as Republican, Aaaaa is the
winner. We now have two candidates to face off in the general; the rest
are eliminated.

What is peculiar about this contrived example is that Ddddd has a higher
approval rating than Aaaaa. However, Ddddd would lose the "instant
primary" for the Democratic nomination to Ccccc in this example.

I've been mulling this over in my head for the past 2-3 days, and it seems
like a way of getting the "top two" aspect of jungle primaries while also
forcing a little diversity via party choice. In this particular example,
the 65 voters that declared themselves either Republicans or Democrats
approve of at least one candidate that advances to the general election
(and additionally, another 5 Libertarians, 5 Socialist, and 5 Green also
approve of one or both of the candidates, for a total of 80 voters seeing
an approved candidate advance). If Ddddd had advanced instead (who had a
higher overall approval), it would have been only 70 voters seeing an
approved candidate advance (since 15 Republicans and 15 Libertarians didn't
approve of either choice).

Turning the party choice into a jungle primary means that Democrats and
Republicans aren't *guaranteed* to get the top two slots. It could mean
than a Democrat and a Socialist get the top two. It could also mean that a
Republican and a Libertarian get the top two.

Does this seem vaguely plausible as a system? Is there some way of gaming
this system I'm not thinking of? Are there plausible scenarios where over
50% disapprove of both candidates who advance? Are there any "false flag"
attacks (i.e. for a "Democrat" to declare themself "Republican" to screw up
their nomination), or would the balancing act be too complicated for most
voters to try?

Rob

[Kevin Venzke]

Hi Rob,
It sounds like under your system parties need to be able to control who can run on their list. I think this moves more of the selection process to prior to the voting. If parties don't want to offer a variety of options under their label then other candidates could be forced to stand as their own party. This might be perfectly viable for them, since you can get one of the top two slots by winning on approval. But if your hope is to get to the final two without being overall approval winner, then you need to run on a list that will get more party votes.
The finalists are selected basically by two separate contests: One finalist is from the list containing the approval winner, and the other from the list with the most party votes. Except, that a list that satisfies both of these has to forfeit one position to the second place party list, so that we can have a second round. I understand why it works that way, and it probably makes the results better, but I think it could seem unfair.. 
Couple of method ideas that occur to me for this setting (no party lists though):1. Round one is an approval ballot. Advance the approval winner, against the candidate with max AO against him. That is, remove the ballots approving the winner and advance, as challenger, whoever is the new approval winner. This doesn't require a first-round auto-win rule (but could support one too, I guess).2. Vote for one favorite and any number of approved. Majority favorite wins. Otherwise, advance the first-preference winner and then, as above, remove ballots that approved this candidate, advancing as challenger the new approval winner.
I guess #1 probably offers more incentive to appeal to the median voter in the first round.
And this next one may be my favorite two-round method these days, though it's probably not as good at addressing concerns about vote-splitting:3. Round one determines by some reasonable method a tentative winner. No one is eliminated. Round two asks whether any of the first-round losers are preferred by a majority to the tentative winner. (Looks basically like an approval ballot, minus one candidate, plus a checkbox to say "none of these are better.")
I would love to have a method that determines a victorious "half" of the voting spectrum in round one, and then in round two lets everyone pick the best candidate from that half. It's hard to see how to make that work though.

Kevin


De : Rob Lanphier <***@robla.net>
À : Election Methods <election-***@lists.electorama.com>
Envoyé le : Jeudi 1 mars 2018 0h25
Objet : [EM] Party-based top two with approval

Hi folks,

The top two jungle primary system we use in California has a lot of folks on edge.  The worry among Democrats and Republicans is that, in some districts, Democrats fear that we'll split the vote and end up with two Republicans.  Republicans fear splitting the vote and ending up with two Democrats in the general election.

One possible alternative that seems more robust to me: for each office, have a ballot as follows:

For Representative from [District X]:

Your preferred party:
[ ] Republican
[ ] Democratic
[ ] Libertarian
[ ] Green
[ ] Socialist

Candidate(s) you approve of:
[ ] Aaaaa (Republican)
[ ] Bbbbb (Republican)
[ ] Ccccc (Democrat)
[ ] Ddddd (Democrat)
[ ] Eeeee (Socialist)
[ ] Ffffff (Libertarian)

The candidate election would be approval based.  The single candidate with the highest approval rating (regardless of party affiliation) would be guaranteed to make it to the general election.  The ballots would also be divided by party preference, such that each party's preferred candidate would be declared the party nominee.  The party with the most votes would be able to choose whether they want to advance their nominee to the general election.  If they opt out (or if their top choice is the same as the overall approval winner), then the next party gets to choose, until a second candidate is chosen. 

As an example, let's say we have 100 people voting with the ballot above.  We end up with this:

Aaaaa  45 votes (by 30 Republicans and 5 Libertarians and 10 Democrats)
Bbbbb  20 votes (by 15 Republicans and 5 Libertarians)
Ccccc  60 votes (by 35 Democrats, 15 Republicans, 5 Socialist, 5 Green)
Ddddd  50 votes (by 30 Democrats, 10 Socialists, 10 Green)
Eeeee  45 votes (by 10 Socialists, 30 Democrats, 5 Green)
Ffffff   30 votes (by 15 Libertarians, 15 Republicans)

The party split:
30 Republican
35 Democratic
15 Libertarian
10 Green
10 Socialist
Ccccc would advance as the general winner.  The Democrats have the most votes, so they have the option of advancing their top vote getter.
Aaaaa 20
Bbbbb 0
Ccccc 35
Ddddd 30
Eeeee 30
Ffffff 0

The self-identified Democrats also pick Ccccc, so we move to the next party.

There are 30 voters who self-identified as Republicans for this election.  Their approved nominee:
Aaaaa 30
Bbbbb 15
Ccccc 15
Ddddd 0
Eeeee 0
Ffffff 15

Among those 30 voters who self-identify as Republican, Aaaaa is the winner.  We now have two candidates to face off in the general; the rest are eliminated.

What is peculiar about this contrived example is that Ddddd has a higher approval rating than Aaaaa.  However, Ddddd would lose the "instant primary" for the Democratic nomination to Ccccc in this example.

I've been mulling this over in my head for the past 2-3 days, and it seems like a way of getting the "top two" aspect of jungle primaries while also forcing a little diversity via party choice.  In this particular example, the 65 voters that declared themselves either Republicans or Democrats approve of at least one candidate that advances to the general election (and additionally, another 5 Libertarians, 5 Socialist, and 5 Green also approve of one or both of the candidates, for a total of 80 voters seeing an approved candidate advance).  If Ddddd had advanced instead (who had a higher overall approval), it would have been only 70 voters seeing an approved candidate advance (since 15 Republicans and 15 Libertarians didn't approve of either choice).

Turning the party choice into a jungle primary means that Democrats and Republicans aren't *guaranteed* to get the top two slots.  It could mean than a Democrat and a Socialist get the top two.  It could also mean that a Republican and a Libertarian get the top two.

Does this seem vaguely plausible as a system?  Is there some way of gaming this system I'm not thinking of?  Are there plausible scenarios where over 50% disapprove of both candidates who advance?  Are there any "false flag" attacks (i.e. for a "Democrat" to declare themself "Republican" to screw up their nomination), or would the balancing act be too complicated for most voters to try?

Rob


----
Election-Methods mailing list - see http://electorama.com/em for list info

[Rob Lanphier]

Hi Kevin,

Thanks for the detailed reply! This is exactly the depth of analysis I was

Post by Kevin Venzke
It sounds like under your system parties need to be able to control who
can run on their list. I think this moves more of the selection process to
prior to the voting.

I was admittedly vague on party affiliation of candidates. The paradox
that's created by the jungle primary system is that a primary elections
have become the preferred way for parties to winnow their candidate lists
to their party's nominee, but parties don't have a way of winnowing their
candidates without having another primary/caucus/whatever.

This problem isn't one that my proposed system introduces. As it stands now
(in California), if I recall correctly, in the current jungle primary
system, it's not very hard to declare oneself as a "Republican" or
"Democrat" in the primary. In a very cursory investigation of the subject,
I see that political party ballot-qualification has a process associated
with it in California at least:
<
http://www.sos.ca.gov/elections/political-parties/political-party-qualification/
I haven't read through that well enough to know all of the existing rules.

It seems to me that a "list" doesn't need to be formally created or
recognized, but *can* organically emerge from whatever candidates run on
the ballot. One possible way to use my proposed system is that the voters
can declare "I'm voting as a Democrat", and then also approve a Socialist
and a Green. If a Socialist or a Green gets the highest approval rating
among self-declared Democrats, then that candidate would be the nominee.
It seems to me that it would be really strategically difficult for a
Republican to win the Democratic "nomination" in this system (or vice
versa), since forcing the voter to identify their party and then identify
the candidates they approve of would make a false flag vote backfire very
easily.

There's a longer reply that I can/should write up to everything else you've
written, but I'll stop myself there for now. Does what I've suggested so
far make sense?

Rob

[Kevin Venzke]

Hi Rob,
I maybe misunderstood part. I would expect in a jungle primary that there wouldn't be a need to restrict what party a candidate claims to be from, since party affiliation isn't a consideration within the method. But under your proposal, parties can decline to advance their winning candidate. So some "party" is making a decision.
I was also thinking (maybe incorrectly on rereading) that the approval winner on each list was always counted from all voters. If that's true then it's important for parties to be able to exclude candidates from their list. If it's not true then there is less need to control who runs as what party. But it seems a little odd if it's not controlled... Taking it to an extreme I can imagine a system where every candidate just has to claim to be left-wing or right-wing and we somehow do something with that.
It sounds like you're saying it's a problem (or else think I view it as a problem) that the parties can't winnow candidates prior to the jungle primary. I don't necessarily view that as a problem... At least, if the jungle primary is supposed to serve as a primary then there shouldn't be a lot of control beforehand. (I'm fine with candidates dropping out on their own when polls say they won't be competitive.) My fear is that if parties have the ability and inclination to winnow candidates prior to the primary, then will they really see a need to offer some choices in the primary?
I guess if under your method, all the party lists just have one candidate, this is hardly broken. An overall Approval winner can run on whatever list they want and advance to the final round.
I would like to deduce party lists organically but it's tricky to come up with something simple. It sounds like maybe you're saying the candidates declare a name for the party they represent and these become options on the ballot for the voters to self-identify. I think that might be abusable... I can imagine a candidate trying to attach themselves (while encouraging the same of their supporters) to a "list" they think they can win, to get an extra boost from the other votes for that list.
Kevin

De : Rob Lanphier <***@robla.net>
À : Kevin Venzke <***@yahoo.fr>
Cc : Election Methods <election-***@lists.electorama.com>
Envoyé le : Vendredi 2 mars 2018 12h46
Objet : Re: [EM] Party-based top two with approval

Hi Kevin,

Thanks for the detailed reply!  This is exactly the depth of analysis I was hoping to get.  One starter reply below:
On Thu, Mar 1, 2018 at 11:34 PM, Kevin Venzke <***@yahoo.fr> wrote:

It sounds like under your system parties need to be able to control who can run on their list. I think this moves more of the selection process to prior to the voting.

I was admittedly vague on party affiliation of candidates.  The paradox that's created by the jungle primary system is that a primary elections have become the preferred way for parties to winnow their candidate lists to their party's nominee, but parties don't have a way of winnowing their candidates without having another primary/caucus/whatever.

This problem isn't one that my proposed system introduces. As it stands now (in California), if I recall correctly, in the current jungle primary system, it's not very hard to declare oneself as a "Republican" or "Democrat" in the primary.  In a very cursory investigation of the subject, I see that political party ballot-qualification has a process associated with it in California at least:
<http://www.sos.ca.gov/elections/political-parties/political-party-qualification/>

I haven't read through that well enough to know all of the existing rules.

It seems to me that a "list" doesn't need to be formally created or recognized, but *can* organically emerge from whatever candidates run on the ballot.  One possible way to use my proposed system is that the voters can declare "I'm voting as a Democrat", and then also approve a Socialist and a Green.  If a Socialist or a Green gets the highest approval rating among self-declared Democrats, then that candidate would be the nominee.  It seems to me that it would be really strategically difficult for a Republican to win the Democratic "nomination" in this system (or vice versa), since forcing the voter to identify their party and then identify the candidates they approve of would make a false flag vote backfire very easily.

There's a longer reply that I can/should write up to everything else you've written, but I'll stop myself there for now. Does what I've suggested so far make sense?

Rob

----
Election-Methods mailing list - see http://electorama.com/em for list info

[Rob Lanphier]

Hi Kevin,

Thanks for thinking this through. I'm struggling to figure out my
position, so it's good to be able to vet this here. More inline...

Post by Kevin Venzke
I maybe misunderstood part. I would expect in a jungle primary that there wouldn't be a need to restrict what party a candidate claims to be from, since party affiliation isn't a consideration within the method. But under your proposal, parties can decline to advance their winning candidate. So some "party" is making a decision.
I was also thinking (maybe incorrectly on rereading) that the approval winner on each list was always counted from all voters. If that's true then it's important for parties to be able to exclude candidates from their list. If it's not true then there is less need to control who runs as what party. But it seems a little odd if it's not controlled... Taking it to an extreme I can imagine a system where every candidate just has to claim to be left-wing or right-wing and we somehow do something with that.

That's kinda where we're at in California today, thanks to Prop 14:

https://en.wikipedia.org/wiki/California_Proposition_14_(2010)

In the status quo, we have a situation which can work out really
poorly for either party. As of right now, there's no good way to
winnow down the list. Right now, both Democrats and Republicans are
trying to get candidates to voluntarily bow out of some races:
http://www.kpbs.org/news/2018/jan/24/both-political-parties-face-uncertain-outcome-49th/

...but they have no orderly way of doing it.

I think, under my proposal, even though candidates would be free to
declare themselves as running under a particular party without support
of that party, it could easily backfire. Since voters declare their
party preference in addition to the candidate, they would need to win
the overall approval vote in order to stay on the ballot. If they
only win the approval of self-declared members of the party, then the
party would have agency to decline their advancement to the general.

Post by Kevin Venzke
It sounds like you're saying it's a problem (or else think I view it as a problem) that the parties can't winnow candidates prior to the jungle primary. I don't necessarily view that as a problem... At least, if the jungle primary is supposed to serve as a primary then there shouldn't be a lot of control beforehand. (I'm fine with candidates dropping out on their own when polls say they won't be competitive.) My fear is that if parties have the ability and inclination to winnow candidates prior to the primary, then will they really see a need to offer some choices in the primary?
I guess if under your method, all the party lists just have one candidate, this is hardly broken. An overall Approval winner can run on whatever list they want and advance to the final round.
I would like to deduce party lists organically but it's tricky to come up with something simple. It sounds like maybe you're saying the candidates declare a name for the party they represent and these become options on the ballot for the voters to self-identify. I think that might be abusable... I can imagine a candidate trying to attach themselves (while encouraging the same of their supporters) to a "list" they think they can win, to get an extra boost from the other votes for that list.

Can you help come up with a more specific example of this? I agree
with you, it *seems* that a candidate would be motivated to attach
themselves to a party in what I referred to as a "false flag"
operation. I fear that you could be right about it being abusable.
However, as it's been rolling around in my head, I haven't been able
to devise a specific example where that actually would actually
happen.

The only false flag operation I could envision being successful is
more of an intra-party insurgency, which seems much more benign. I
think the idea of a party sincerely getting co-opted by voters willing
to self-identify as supporting that party can be a very good thing.

Rob
----
Election-Methods mailing list - see http://electorama.com/em for list info

[Kevin Venzke]

Just a quick idea on the last point: I can imagine a situation where there's a minor, established party called "Reform," and in some election an independent candidate decides he's going to label himself "Reform" because 1. the label isn't too much of a stretch, 2. he thinks he's strong enough to win within that list, and 3. he wants a boost from the "established" Reform voters. If the original Reform voters aren't OK with this, they don't have much recourse. They can change their name, but they can't protect the new name either. Whether it's "false flag" would be disputed.
But this is assuming there is no actual party with a role in the process. Below it sounds like you're saying there is still a Reform party who can decline to advance the candidate. So I can declare that I'm running as Reform, but it means that *if I win* on the list, the Reform party organization can opt to throw Reform's win away and let someone else have it. That's an interesting idea. The party still has control of the list, but only at the end, and always at the expense of forfeiting their "win."
I wonder if such a system would be "stable," as in, have general acceptance. A party in the position of wanting to forfeit their seats because of rogue candidates trying to piggyback, I can imagine would start to argue for a change in the rules. On the other hand if a seat is forfeited it's presumably lose-lose for the party, voters, and the candidate. So maybe it's a threat that would never actually happen.
Kevin 

De : Rob Lanphier <***@robla.net>
À : Kevin Venzke <***@yahoo.fr>
Cc : Election Methods <election-***@lists.electorama.com>
Envoyé le : Lundi 5 mars 2018 0h04
Objet : Re: [EM] Party-based top two with approval

Hi Kevin,

Thanks for thinking this through.  I'm struggling to figure out my
position, so it's good to be able to vet this here.  More inline...

Post by Kevin Venzke
I maybe misunderstood part. I would expect in a jungle primary that there wouldn't be a need to restrict what party a candidate claims to be from, since party affiliation isn't a consideration within the method. But under your proposal, parties can decline to advance their winning candidate. So some "party" is making a decision.
I was also thinking (maybe incorrectly on rereading) that the approval winner on each list was always counted from all voters. If that's true then it's important for parties to be able to exclude candidates from their list. If it's not true then there is less need to control who runs as what party. But it seems a little odd if it's not controlled... Taking it to an extreme I can imagine a system where every candidate just has to claim to be left-wing or right-wing and we somehow do something with that.

That's kinda where we're at in California today, thanks to Prop 14:

https://en.wikipedia.org/wiki/California_Proposition_14_(2010)

In the status quo, we have a situation which can work out really
poorly for either party.  As of right now, there's no good way to
winnow down the list.  Right now, both Democrats and Republicans are
trying to get candidates to voluntarily bow out of some races:
http://www.kpbs.org/news/2018/jan/24/both-political-parties-face-uncertain-outcome-49th/

...but they have no orderly way of doing it.

I think, under my proposal, even though candidates would be free to
declare themselves as running under a particular party without support
of that party, it could easily backfire.  Since voters declare their
party preference in addition to the candidate, they would need to win
the overall approval vote in order to stay on the ballot.  If they
only win the approval of self-declared members of the party, then the
party would have agency to decline their advancement to the general.

Post by Kevin Venzke
It sounds like you're saying it's a problem (or else think I view it as a problem) that the parties can't winnow candidates prior to the jungle primary. I don't necessarily view that as a problem... At least, if the jungle primary is supposed to serve as a primary then there shouldn't be a lot of control beforehand. (I'm fine with candidates dropping out on their own when polls say they won't be competitive.) My fear is that if parties have the ability and inclination to winnow candidates prior to the primary, then will they really see a need to offer some choices in the primary?
I guess if under your method, all the party lists just have one candidate, this is hardly broken. An overall Approval winner can run on whatever list they want and advance to the final round.
I would like to deduce party lists organically but it's tricky to come up with something simple. It sounds like maybe you're saying the candidates declare a name for the party they represent and these become options on the ballot for the voters to self-identify. I think that might be abusable... I can imagine a candidate trying to attach themselves (while encouraging the same of their supporters) to a "list" they think they can win, to get an extra boost from the other votes for that list.

Can you help come up with a more specific example of this?  I agree
with you, it *seems* that a candidate would be motivated to attach
themselves to a party in what I referred to as a "false flag"
operation.  I fear that you could be right about it being abusable.
However, as it's been rolling around in my head, I haven't been able
to devise a specific example where that actually would actually
happen.

The only false flag operation I could envision being successful is
more of an intra-party insurgency, which seems much more benign.  I
think the idea of a party sincerely getting co-opted by voters willing
to self-identify as supporting that party can be a very good thing.

Rob

[Rob Lanphier]

Hi Kevin (and everyone)

Thanks for engaging on this topic. I think I've become convinced that
an approval-based solution is workable. I don't know if I can defend
the specific rules that I laid out, but I believe that a big,
multi-candidate (and multi-party) primary, and a top-two general is a
good combination. I think there's viable opportunity for reform within
those constraints.

I'll keep exploring the specific set of rules that I laid out in this
thread. However, I'm not so sure that my specific rules are the best
reform. My rules are a little too tough to explain. More inline,
though....

Post by Kevin Venzke
Just a quick idea on the last point: I can imagine a situation where there's
a minor, established party called "Reform," and in some election an
independent candidate decides he's going to label himself "Reform" because
1. the label isn't too much of a stretch, 2. he thinks he's strong enough to
win within that list, and 3. he wants a boost from the "established" Reform
voters. If the original Reform voters aren't OK with this, they don't have
much recourse. They can change their name, but they can't protect the new
name either. Whether it's "false flag" would be disputed.

That's a really interesting point. That's more-or-less what happened
with the Reform party in the U.S. In 1992 and 1996, it was
effectively the Perot party. In 2000, Pat Buchanan won the Reform
party nomination, which wouldn't have been the first time he engaged
in subterfuge of parties opposed to Republicans[1]

[1]: https://www.washingtonpost.com/wp-srv/national/longterm/watergate/stories/buchananmemo.htm

Post by Kevin Venzke
But this is assuming there is no actual party with a role in the process.
Below it sounds like you're saying there is still a Reform party who can
decline to advance the candidate. So I can declare that I'm running as
Reform, but it means that *if I win* on the list, the Reform party
organization can opt to throw Reform's win away and let someone else have
it. That's an interesting idea. The party still has control of the list, but
only at the end, and always at the expense of forfeiting their "win."

One way that it could work: the *overall* approval vote winner is
automatically on the ballot (regardless of whether they the winning
party or not). But then, the top vote getting party would have the
opportunity to put a candidate on the ballot if their voters made a
different choice. Let me try tweaking my example. Once again, 100
people voting. I'll propose a tweak to my example ballot:

Your preferred party (mark only one)
[ ] Republican
[ ] Democratic
[ ] Libertarian
[ ] Green
[ ] Socialist
[ ] (no party preference)

Candidate(s) you approve of (mark multiple candidates if you desire):
[ ] Aaaaa (Republican)
[ ] Bbbbb (Republican)
[ ] Ccccc (Democrat)
[ ] Ddddd (Democrat)
[ ] Eeeee (Socialist)
[ ] Ffffff (Libertarian)

In both my original example, and the example below, I say something like:

Aaaaa 45 votes (by 30 Republicans and 5 Libertarians and 10 Democrats)

What that means is that for the "10 Democrats", this is their vote:
Your preferred party:
[X] Democratic

Candidate(s) you approve of:
[X] Aaaaa (Republican)
(and any other candidates this voter approves of)

I'll suggest a slightly different mix of voters:
Aaaaa 45 votes (by 30 Republicans and 5 Libertarians and 10 Democrats)
Bbbbb 20 votes (by 15 Republicans and 5 Libertarians)
Ccccc 60 votes (by 29 Democrats, 11 Republicans, 5 Socialist, 5
Green, 10 no party preference)
Ddddd 50 votes (by 29 Democrats, 10 Socialists, 10 Green, 1 no preference)
Eeeee 45 votes (by 10 Socialists, 30 Democrats, 5 Green)
Ffffff 30 votes (by 15 Libertarians, 15 Republicans)1

The party split:
25 Republican
30 Democratic
15 Libertarian
10 Green
10 Socialist
10 no party preference

As before, Ccccc would advance as the general winner. In this example
(as with my initial example in my 2018-02-28 email), the Democrats
have the most votes, so they have the option of advancing their top
vote getter. The difference, though, is that their top Democratic
vote getter isn't Ccccc:

Aaaaa 20
Bbbbb 0
Ccccc 29
Ddddd 29
Eeeee 30
Ffffff 0

The self-identified Democrats pick Eeeee.

Note that in this example, Ccccc has the highest overall approval.
However, Eeeee only has the highest approval among voters who voted
for the Democratic party. Ddddd has higher overall approval than
Eeeee (50-45), but didn't earn the highest Democratic approval.
Moreover, Eeeee merely ties Aaaaa in overall approval. However, the
Democratic party still secures a much higher share of the party vote
than the Republican party, and thus earns the right to advance their
preferred candidate (even though Eeeee is listed as "Socialist").
Thus the general election would be between Ccccc and Eeeee.

Voters who identify as Republicans would likely be furious about two
"Democrats" advancing (both Ccccc the declared Democrat, and Eeeee,
the Socialist nominated by the Democrats). In fairness, though, Ccccc
(in this example) appears to be the centrist who did the best job of
running a bi-partisan campaign (where "bi-partisan" refers to
Democrats and Republicans, and ignores the others in proud American
tradition). Our fictional Republicans in this example can instead
choose to be upset with the 11 "Republican" voters who cast their
approval of Ccccc.

I think this particular example ends well for Democrats. However, I
don't think there is a way that Democrats (or Republicans, or a major
party) to reliably strategize to win both spots.

Note that I nearly changed this example to make it so that Ccccc and
Ddddd were the two candidates. It's not hard to come up with an
example where the two declared Democrats advance in this system,
though only one candidate would be a party nominee. Ironically (in my
Ccccc vs Eeeee example), Eeeee would be the Democratic nominee, and
would be the only candidate that the Democratic party could decide to
forfeit. That would then allow the Republicans to advance their
nominee (Aaaaa).

I suspect that a much simpler version that doesn't involve parties at
all might be superior. Kevin's earlier suggestion looks really

Post by Kevin Venzke
1. Round one is an approval ballot. Advance the approval winner, against the
candidate with max AO against him. That is, remove the ballots approving the
winner and advance, as challenger, whoever is the new approval winner. This
doesn't require a first-round auto-win rule (but could support one too, I guess).

The thing I like about this first system Kevin proposes is that it
seems reasonably likely to satisfy this criteria: two candidates for
whom the largest proportion of voters approves of at least one of the
two candidates. Kevin's system #1 has the benefit of simplicity. A
potential downside of that system is that having a popular leading
candidate means that many fewer voters are involved in choosing the
runner up (e.g. if the top candidate gets 85% approval, then the
runner up is chosen by the 15% that didn't like the top candidate).

Post by Kevin Venzke
I wonder if such a system would be "stable," as in, have general acceptance.
A party in the position of wanting to forfeit their seats because of rogue
candidates trying to piggyback, I can imagine would start to argue for a
change in the rules. On the other hand if a seat is forfeited it's
presumably lose-lose for the party, voters, and the candidate. So maybe it's
a threat that would never actually happen.

An example of where the party could decide to forfeit could be a
situation similar to the one Republicans face in Illinois:
<https://www.theguardian.com/us-news/2018/feb/04/neo-nazi-holocaust-denier-art-jones-illinois-republicans>

That particular example is one where the Republicans claim they want
to jettison the candidate, but don't have the means to do it.

One thing that I think makes my system resilient against a party
forfeiting a popular candidate is that the party could be held to
account in a subsequent election. Forfeiting a popular candidate
would make it far less likely for the voters who voted for that party
to vote for that party again.

Rob
----
Election-Methods mailing list - see http://electorama.com/em for list info

[Kevin Venzke]

Hi Rob,

Post by Rob Lanphier
One way that it could work: the *overall* approval vote winner is
automatically on the ballot (regardless of whether they the winning
party or not).  But then, the top vote getting party would have the
opportunity to put a candidate on the ballot if their voters made a
different choice.  Let me try tweaking my example.  Once again, 100

Well, what bothers me is that it can happen that a Democrat is the overall approvalwinner while the Democratic party list is the party list winner, simply because theirrivals are divided a bit. If you have only Democrats advance in that scenario then itseems like the two rounds have less purpose.
I think the design should make sure that the same voters can't supply both finalists,but without raising the question of whether a party can do "so well" that we have toforcibly take one of the selections away from them.

Post by Rob Lanphier
I suspect that a much simpler version that doesn't involve parties at
all might be superior.  Kevin's earlier suggestion looks really

Post by Kevin Venzke
1. Round one is an approval ballot. Advance the approval winner, against the
candidate with max AO against him. That is, remove the ballots approving the
winner and advance, as challenger, whoever is the new approval winner. This
doesn't require a first-round auto-win rule (but could support one too, I guess).

The thing I like about this first system Kevin proposes is that it
seems reasonably likely to satisfy this criteria: two candidates for
whom the largest proportion of voters approves of at least one of the
two candidates.  Kevin's system #1 has the benefit of simplicity.  A
potential downside of that system is that having a popular leading
candidate means that many fewer voters are involved in choosing the
runner up (e.g. if the top candidate gets 85% approval, then the
runner up is chosen by the 15% that didn't like the top candidate).

Yes, though if it's 85% I think the unfortunate thing might be that we would haveto go on to have a second round at all. I think if a candidate can get majority approval (especially on a ballot without ranks) the biggest concern is addressed.
Maybe stating the obvious but you could if you want directly advance the twocandidates who maximize the number of voters who approved at least one. A majordownside of doing that would be that the overall approval winner isn't guaranteedto be one of the finalists.
There are some interesting questions about what we're trying to do when pickingfinalists. The overall approval winner is a good estimate for the best overall candidate, so it seems bad to omit him. The two candidates who "maximizeparticipation" (let's say) is more like an estimate of who the voters think are the main contenders.
Regarding the 85%/15% scenario I almost want to say "use proportional approval voting so the 85% can contribute somewhat in picking the second finalist aswell." That would probably give you a more competitive second round.
Without getting into proportionality (which isn't directly related to the goal of thefirst round) one could probably come up with an equation that balances eachcandidate's likelihood of being able to beat the approval winner pairwise, withtheir likelihood of simply being a clone of the approval winner.

Post by Rob Lanphier

Post by Kevin Venzke
I wonder if such a system would be "stable," as in, have general acceptance.
A party in the position of wanting to forfeit their seats because of rogue
candidates trying to piggyback, I can imagine would start to argue for a
change in the rules. On the other hand if a seat is forfeited it's
presumably lose-lose for the party, voters, and the candidate. So maybe it's
a threat that would never actually happen.

An example of where the party could decide to forfeit could be a
That particular example is one where the Republicans claim they want
to jettison the candidate, but don't have the means to do it.
One thing that I think makes my system resilient against a party
forfeiting a popular candidate is that the party could be held to
account in a subsequent election.  Forfeiting a popular candidate
would make it far less likely for the voters who voted for that party
to vote for that party again.

I would only see forfeits occurring if the party views the candidate's supportersas "invasive" in the first place, so they wouldn't be concerned about losing thesupport. I can't see a party forfeiting any candidate who is at least somewhat appropriate for the list.
Kevin