Polling Ourselves

Lime

What's the goal with such a poll? Is it to provide information for the public, or to make some decision within this group?

Ideally I'd like to see a poll including as many economists, social choice theorists, voting reform advocates, etc. as possible with snowball sampling, and publishing all ballots together with a data analysis (rather than just a single winner or ordering). The results can be published in a paper that can be cited to support claims like "Experts generally believe system A is better than B."

Jack Waugh

@lime There is no proposal to make any formal decision within the group.

Some have proposed repeated polling and I agree with this.

I think the goal of the current one is to inform us of our aggregate attitude toward the single-winner systems, mainly as to their merits for proposing to the public for political elections. Knowing this attitude and some statistics over our votes (do they cluster?) might lead to discussions that we might otherwise not have thought to enter. I think this is valuable. Also, some of us like to describe or invent systems as intellectual curiosities, and the polls (including the current one) can help us as a group distinguish between those and the serious proposals.

I would not oppose trying to implement for a future polling, the ideal you state. You are not the first to suggest asking for votes from members of the larger voting-systems interest community going beyond just the participants in this forum.

Jack Waugh

@lime The public can read this forum and if any of us thinks the set of participants in the current polling or a future one merits mentioning the poll and its outcome in other fora, antisocial media, and/or personal communications, I don't see anything wrong with that. Promoting reform or revolution can be a goal.

@toby-pereira said in Polling Ourselves:

@michaelossipoff said in Polling Ourselves:

Wouldn’t Score(0-100) be better than Score(0-9) for the RP/Score hybrid? It would have room for expression of all pairwise preferences, instead of sometimes forcing equal-rating.

Possibly. I kept it at 0-9 for simplicity, but I wouldn't object to 0-100 or 0-99. (There has been debate in the past over whether 10 and 100 or 9 and 99 are better as max scores. I'm fairly unbothered by it.)

@toby-pereira said in Polling Ourselves:

@michaelossipoff said in Polling Ourselves:

Wouldn’t Score(0-100) be better than Score(0-9) for the RP/Score hybrid? It would have room for expression of all pairwise preferences, instead of sometimes forcing equal-rating.

Possibly. I kept it at 0-9 for simplicity, but I wouldn't object to 0-100 or 0-99. (There has been debate in the past over whether 10 and 100 or 9 and 99 are better as max scores. I'm fairly unbothered by it.)

0-9 or 0-99 is much more line space-efficient. 0-10 or 0-100 gives you one more rating-slot, , 10/9 or 100/99 times as many rating-slots, but requires 2 or 3/2 times as much linespace.

But linespace isn’t an issue at all in our balloting, & so, if it were my choice, I’d use 0-10 or 0-100.

But either is fine with me.

But you’re probably right to use 0-9 instead of 0-99 for Hybrid, because someone could rate 99 & 98, in order to effectively do all-or-nothing strategy while still expressing a pairwise preference.

Hybrid doesn’t combine pairwise expression with sincerity incentive as well as I thought, but I’ll still rank & rate it as high as Score(100, 99, 50, 1, 0).

I heard Williamson in a debate, & based on that, I’ll still rank her 2nd, but I won’t approve her.

It’s taking me this long to get to re-posting my poll-ballots because I choose to vote every by every method proposed for the poll. A few others likely will also, but I emphasize that it isn’t necessary for participation.

You can, for example, just vote an Approval-set & a ranking, & maybe a Score ballot by your 1 favorite Score balloting.

(My favorite Score balloting is now Score(0-5). )

The tacitly-agreed-upon end-date for voting is still 10 days away.

Anyway, my ballots:

Voting-Systems:

Approval Set:

{Approval, RP(wv)}

Ranking:

Approval
RP(wv)
Score(0-5)
Score(0-10)
Score(100, 99, 90, 50, 10, 1, 0)
Hybrid
Score(2,1, 0)

(I think that’s all of the “ candidates”.)

By Score(0-5) (for elections):

Approval: 5
RP(wv): 5
Score(0-5): 3
Score(100, 99, 90, 50, 10, 1, 0): 2
Score(0-10) & Hybrid : 1

By Score(0-5) (for polls):

RP(wv), with Score(0-5) & Approval, & maybe Score(100, 99, 90, 50, 10, 1, 0): 5

By Score(100, 99, 90, 50, 10, 1, 0) (elections):

Approval: 100
RP(wv): 100
Score(0-5): 50
Score(100, 99, 90, 50, 10, 1, 0): 50
Score(0-10): 10
Score (2, 1, 0): 1

By Score(100, 99, 90, 50, 10, 1, 0) (pols):

RP(wv), with Approval, Score(0-5), & maybe Score(100, 99, 90, 50, 10, 1, 0): 100

Presidential Candidates:

Approval Set:

{Jill Stein}

Ranking:

Jill Stein
Marianne Williamson

By Score(0-5):

Jill Stein: 5
Anyone else: 0

Jack Waugh

@michaelossipoff said in Polling Ourselves:

(I think that’s all of the “ candidates”.)

They are at https://www.votingtheory.org/forum/post/3169 .

Hybrid

I suppose by that you might mean any and all of:

STAR (Ossipoff, Waugh)
Smith//Score (0 to 9 ballot) (Pereira)
quantile-normalized score, with integer scores from 0 to 100 (Frankston)

Jack Waugh

Preliminary result.

Approval shows a commanding lead!

I edited and may edit again.
Fri Mar 15 06:34:20 AM UTC 2024
For elections.
Based on Score{100, 99, 90, 50, 10, 1, 0}.

200 Approval
150 Score{5, 4, 3, 2, 1, 0}
149 Score{100, 99, 90, 50, 10, 1, 0}
110 0-9 score [note 0]
101 Score{2, 1, 0}
101 0 to 9 scale (only used for ranking),
    ranked pairs, winning votes. [note 1]
100 Ranked-Pairs(winning-votes) equal-ranking
    allowed
099 STAR
000 quantile-normalized score, with integer
    scores from 0 to 100

[0] @MichaelOssipoff says Score(0-10), which is not nominated, and I am counting it as Score{0-9}.
[1] Ossipoff does not directly indicate this nominee, but I'm assuming he likes it as much as he does the ranking version allowing equal ranking, since for purposes of determining the winner, the scores in this are only an encoding of a ranking allowing equal-ranking.

By “Hybrid”, i just meant Score(0-9) balloting counted as RP(wv).

Yes, Approval is the natural compromise, though we like different elaborations.

I forgot about STAR. I rank & rate it with Hybrid.

Jack Waugh

@michaelossipoff said in End of poll-balloting at the next midnight:

Score(100, 99, 90, 50, 10, 1, 0)

Your and my ballots are the only ones I know about. I tried to tally them in Score(100, 99, 90, 50, 10, 1, 0), but I based that version of the tally on an interpretation of your ballot that turned out to be not quite in accord with your later clarification. But the lead that that tallying shows for Approval over the rest of the pack is probably pretty accurate.

Jack Waugh

I don't know how to tally in RP(wv).

For purposes of tallying

Score(0-5)
Score(100, 99, 90, 50, 10, 1, 0),

here is an attempt to interpret @MichaelOssipoff's ballot while staying confined to the actual nominees.

First, let's list again the nominees as I had compiled above.

Approval
Ranked-Pairs(winning-votes) equal-ranking allowed
STAR
Score{2, 1, 0}
Score{100, 99, 90, 50, 10, 1, 0}
Score{5, 4, 3, 2, 1, 0}
0 to 9 scale (only used for ranking), ranked pairs, winning votes
Smith//Score (0 to 9 ballot)
quantile-normalized score, with integer scores from 0 to 100
0-9 score

Now let's bring in @MichaelOssipoff's "ballot" as of the most recent place where he tries to present it in full:

@michaelossipoff said in Polling Ourselves:

By Score(0-5) (for elections):
Approval: 5
RP(wv): 5
Score(0-5): 3
Score(100, 99, 90, 50, 10, 1, 0): 2
Score(0-10) & Hybrid : 1

... [skipping entries "for polls", a category we never agreed on]

By Score(100, 99, 90, 50, 10, 1, 0) (elections):
Approval: 100
RP(wv): 100
Score(0-5): 50
Score(100, 99, 90, 50, 10, 1, 0): 50
Score(0-10): 10
Score (2, 1, 0): 1

He later explains what "Hybrid" means:

@michaelossipoff said in Polling Ourselves:

By “Hybrid”, i just meant Score(0-9) balloting counted as RP(wv).
...
I forgot about STAR. I rank & rate it with Hybrid.

So that explains Hybrid.

The other problem with Michael's ballot is that it mentions Score(0-10), a non-nominee. In my preliminary tally above, I counted that as 0-9 score. But this time, I'm in the mood to be strict (since this is the end), so I am going to throw it out.

So let's map the entries of his Score(0-5) ballot to the nominees:

5 Approval
5 Ranked-Pairs(winning-votes) equal-ranking allowed
1 STAR
0 Score{2, 1, 0}
2 Score{100, 99, 90, 50, 10, 1, 0}
3 Score{5, 4, 3, 2, 1, 0}
1 0 to 9 scale (only used for ranking), ranked pairs, winning votes
0 Smith//Score (0 to 9 ballot)
0 quantile-normalized score, with integer scores from 0 to 100
0 0-9 score

And let's do the same wth his Score(100, 99, 90, 50, 10, 1, 0) ballot:

100 Approval
100 Ranked-Pairs(winning-votes) equal-ranking allowed
000 STAR
001 Score{2, 1, 0}
050 Score{100, 99, 90, 50, 10, 1, 0}
050 Score{5, 4, 3, 2, 1, 0}
000 0 to 9 scale (only used for ranking), ranked pairs, winning votes
000 Smith//Score (0 to 9 ballot)
000 quantile-normalized score, with integer scores from 0 to 100
000 0-9 score

Jack Waugh

Tally in 0-5:

(Ossipoff + Waugh = total)

same order as I listed the nominees

5 + 5 = 10 Approval
5 + 0 = 05 Ranked-Pairs(winning-votes) equal-ranking allowed
1 + 4 = 05 STAR
0 + 5 = 05 Score{2, 1, 0}
2 + 5 = 07 Score{100, 99, 90, 50, 10, 1, 0}
3 + 5 = 08 Score{5, 4, 3, 2, 1, 0}
1 + 1 = 02 0 to 9 scale (only used for ranking), ranked pairs, winning votes
0 + 4 = 04 Smith//Score (0 to 9 ballot)
0 + 0 = 00 quantile-normalized score, with integer scores from 0 to 100
0 + 5 = 05 0-9 score

Reorder by descending totals:

5 + 5 = 10 Approval
3 + 5 = 08 Score{5, 4, 3, 2, 1, 0}
2 + 5 = 07 Score{100, 99, 90, 50, 10, 1, 0}
5 + 0 = 05 Ranked-Pairs(winning-votes) equal-ranking allowed
1 + 4 = 05 STAR
0 + 5 = 05 Score{2, 1, 0}
0 + 5 = 05 0-9 score
0 + 4 = 04 Smith//Score (0 to 9 ballot)
1 + 1 = 02 0 to 9 scale (only used for ranking), ranked pairs, winning votes
0 + 0 = 00 quantile-normalized score, with integer scores from 0 to 100

Approval has a 20% lead over the runner-up.

Jack Waugh

Tally in Score{100, 99, 90, 50, 10, 1, 0}:

(Ossipoff + Waugh = total)

same order as I listed the nominees

100 + 100 = 200 Approval
100 + 000 = 100 Ranked-Pairs(winning-votes) equal-ranking allowed
000 + 099 = 099 STAR
001 + 100 = 101 Score{2, 1, 0}
050 + 099 = 149 Score{100, 99, 90, 50, 10, 1, 0}
050 + 100 = 150 Score{5, 4, 3, 2, 1, 0}
000 + 001 = 001 0 to 9 scale (only used for ranking), ranked pairs, winning votes
000 + 090 = 090 Smith//Score (0 to 9 ballot)
000 + 000 = 000 quantile-normalized score, with integer scores from 0 to 100
000 + 100 = 100 0-9 score

Reorder by descending totals:

100 + 100 = 200 Approval
050 + 100 = 150 Score{5, 4, 3, 2, 1, 0}
050 + 099 = 149 Score{100, 99, 90, 50, 10, 1, 0}
001 + 100 = 101 Score{2, 1, 0}
100 + 000 = 100 Ranked-Pairs(winning-votes) equal-ranking allowed
000 + 100 = 100 0-9 score
000 + 099 = 099 STAR
000 + 090 = 090 Smith//Score (0 to 9 ballot)
000 + 001 = 001 0 to 9 scale (only used for ranking), ranked pairs, winning votes
000 + 000 = 000 quantile-normalized score, with integer scores from 0 to 100

Approval beats the runner-up by 25%, he exclaimed.