Fair redistricting: California

[President Punxsutawney Phil]

I'll start building a table to post SPICE scores for NM submissions here. NM has a PVI of D+3 and that projects to an advantage of 0.36 of a seat for the Dems. Rounding to whole numbers that would require a plan with 1 D 1R and 1 even seat to have a SKEW of 0. I think all the plans are 2 D and 1 R and they will have a SKEW of 1. The Albuquerque UCC is Bernalillo, Sandoval and Valencia and if those counties have more than 2 CDs it counts as an extra chop, as it does if no CD is nested in those three counties. A plan that picks up just the precinct needed to keep a reservation whole does not count as a chop, but if it picks up additional precincts in the chopped county it does count as a chop.

Plan	-S-	-P-	-I-	-C-	-E-
HCP	1	6	5	4	12
Sol	1	6	3	1	14
Solid4096	1	6	5	2	20
TimTurner	1	6	11	2	12
jimrtex	1	6	5	2	20

Jimrtex's whole county plan is the same as Solid4096.

If Sol is actually getting better scores with his very flawed plan than other plans that actually use sensible design methodology, then I think there is something that needs to be fixed with the muon rules. Probably the underlined part.

The muon rules are designed that way as to encourage keeping CDs within the cores of metropolitan areas; something that isn't exactly a bad objective to seek.

[muon2]

I'll start building a table to post SPICE scores for NM submissions here. NM has a PVI of D+3 and that projects to an advantage of 0.36 of a seat for the Dems. Rounding to whole numbers that would require a plan with 1 D 1R and 1 even seat to have a SKEW of 0. I think all the plans are 2 D and 1 R and they will have a SKEW of 1. The Albuquerque UCC is Bernalillo, Sandoval and Valencia and if those counties have more than 2 CDs it counts as an extra chop, as it does if no CD is nested in those three counties. A plan that picks up just the precinct needed to keep a reservation whole does not count as a chop, but if it picks up additional precincts in the chopped county it does count as a chop.

Plan	-S-	-P-	-I-	-C-	-E-
HCP	1	6	5	4	12
Sol	1	6	3	1	14
Solid4096	1	6	5	2	20
TimTurner	1	6	11	2	12
jimrtex	1	6	5	2	20

Jimrtex's whole county plan is the same as Solid4096.

If Sol is actually getting better scores with his very flawed plan than other plans that actually use sensible design methodology, then I think there is something that needs to be fixed with the muon rules. Probably the underlined part.

One thing that helps Sol is the special rule for reservations. It exempts his plan from a chop of Bernalillo, and reduces his plan erosity by 1. Otherwise his plan would be comparable to mine, since I didn't use the reservation exemption. The UCC rules arose to give additional weight to plans that preserved a district for metro areas and didn't dilute rural areas with metro counties. They are intended to prefer maps that keep a metro intact compared to a whole county plan, if everything else is equal.

[muon2]

I think I have scored all the plans. Let me know if I missed one.

Plan	-S-	-P-	-I-	-C-	-E-
HCP	1	6	5	4	12
Sol	1	6	3	1	14
Solid4096	1	6	5	2	20
omegascarlet	1	6	6	6	16
TimTurner	1	6	11	2	12
muon2	1	6	3	1	15
AutralianSwingVoter	1	6	5	2	14
singletxguyforfun	1	5	5	4	19
jimrtex	1	6	5	2	20

Jimrtex's whole county plan is the same as Solid4096. The score for his plan that shifts one precinct from Sandoval to Bernalillo increases the chop and erosity by 1 each and decreases the inequality by 2: (1, 6, 3, 3, 21).

[jimrtex]

Jimrtex's whole county plan is the same as Solid4096. The score for his plan that shifts one precinct from Sandoval to Bernalillo increases the chop and erosity by 1 each and decreases the inequality by 2: (1, 6, 3, 3, 21).

I intended to also shift the portion of Valencia.

[muon2]

Jimrtex's whole county plan is the same as Solid4096. The score for his plan that shifts one precinct from Sandoval to Bernalillo increases the chop and erosity by 1 each and decreases the inequality by 2: (1, 6, 3, 3, 21).

I intended to also shift the portion of Valencia.

I understand, but I can't replicate a map that chops a precinct. Since DRA is the tool of choice for this exercise, I am constrained in that way. The way you've drawn it it appears to increase chop by 1, erosity by 1, and decrease inequality by 2.

[Pennsylvania Deplorable]

Here's my New Mexico proposal. Sandoval and Bernallillo are the only split counties; Bernallillo is split along reservation lines.




The logic was to keep the Albequerque metro in as few districts as possible, and to unite as much of the Native groups in the NW as possible. It turned out that the best way to accomplish both was by having NM-01 donut hole around the city. NM-01 is plurality Latinx by population but plurality White by VAP, Natives are a strong minority.

NM-01: Dv. -119, PVI D+8.49
NM-02: Dv. +66, PVI R+7.01
NM-03: Dv. +53, PVI D+7.15

I like this map. The third is nice and compact, there are barely any county splits, and it works well with the cultural divide (NM-2 being an extension of west Texas culturally)

[Sol]

Since the time is up on New Mexico, I figured I might would give Colorado a try. It's a surprisingly hard state to redistrict; there's a lot of weird constraints thanks to roads and mountain passes. My fifth district actually is not contiguous by road as Saguache is blocked by mountains, but the road goes just slightly through Chaffee so I figure it wouldn't be a huge impediment. I might submit a map which disregards UCCs and county splits more in favor of communities of interest, since there are some things on this map which I don't love.




After I finished tinkering with these maps I realized I was over my acceptable threshold in the fourth district for deviation and fixed it after taking the jpegs. As a result there's a couple precinct's difference in my analysis below from the map above, namely in the split of Douglas County (and one precinct different in the split of Weld). Makes very little difference in PVI or district shape but I can give that if y'all want it.

CO-01: Dv. -531, PVI D+24.91
CO-02: Dv. -59, PVI D+8.97
CO-03: Dv. -804, PVI R+4.99
CO-04: Dv. 789, PVI R+13.04
CO-05: Dv. 549, PVI R+13.75
CO-06: Dv. -536, PVI D+1.4
CO-07: Dv. +254, PVI D+4.65

So Democrats and Republicans both get 2 safe seats, with CO-07 and CO-03 being likely D and R respectively. C0-06 is a straight-up tossup.

[Strudelcutie4427]

My votes on NM, Ive been gone all weekend. Has a winner been picked yet?

AutralianSwingVoter
HCP
Muon
Sol
Singletxguy
TimTurner
Omega
Jimrtex
Solid

Also here is my Colorado
https://snag.gy/GZwE1l.jpg

1. D+25 (-842)
2. D+9 (-1575)
3. R+8 (1452)
4. R+7 (664)
5. R+11 (-736)
6. R+5 (-294)
7. D+5 (1328)

All County chops were done for population and done in a way that keeps more of a compact overall district shape
1. takes in West Aurora
2. takes in Western Weld
3. none
4. 3 chops. East Adams, East Arapahoe, Southern Pueblo
5. Northern Pueblo
6, Southern Arapahoe
7. Adams-Jeffco split

[cvparty]

HOLD UP GUYS DON'T MOVE ON TO THE NEXT STATE

[jimrtex]

No County Splits. District 2 includes 78% of AIAN population.

NM-1 (Albuquerque) -0.11%; D 7.33; H 48%, A 42%; AIAN 4%; B 1%; As 2%; O 2%.
NM-2 (North) +0.09%; D 9.49; H 39%; A 38%; AIAN 20%; O 2%; B 1%; As 1%.
NM-3 (South) +0.02%; R 8.86; H 52%; A 42%: AIAN 2%; B 2%; As 1%; O 1%.

Since my whole county map was the same as Solid4096, this is an alternative that keeps the Albuquerque UCC in two districts.



NM-1 (Albuquerque) -0.11%; D 7.33; H 48%, A 42%; AIAN 4%; B 1%; As 2%; O 2%.
NM-2 (North) -0.07%; D 8.34; H 39%; A 38%; AIAN 20%; O 2%; As 1%; B 1%.
NM-3 (South) +0.17%; R 7.02; H 53%; A 42%; B 2%; AIAN 2%; O 1%; As 1%.

Standard deviation 0.12%. 79% of AIAN population is in NM-2.

If you want a hideous looking map, you can shift Harding to NM-2 and reduce the standard deviation to 0.08%.

[cvparty]

SOL'S MAP WINS. We're now on Colorado! Please provide a Denver close-up image in addition to your image. Like so (this is my submission)


The satellite image is to show/justify my Jeffco split, you can see the sparsely populated mountains serving as the 6th's and 7th's western boundary
1: D+23
2: D+10
3: R+6
4: R+15
5: R+14
6: D+3
7: D+5

[jimrtex]

SOL'S MAP WINS.

Why?

[cvparty]

SOL'S MAP WINS.

Why?

Because people ranked it highly

[jimrtex]

SOL'S MAP WINS.

Why?

Because people ranked it highly

Has anyone other than you seen the rankings?

[cvparty]

Here are the votes:
TimTurner: (not sure if the order or the numbers indicate the rank, but either way Sol's would win)
5. HCP
1. Sol
6. Solid4096/jimrtex
4. Scarlet
2. TimTurner
3. muon2
7. AustralianSwingVoter
8. Singletxguyforfun

Sol: Sol>muon2=Scarlet>AustralianSwingVoter>Solid>HCP=TimTurner
cvparty: scarlet>muon2>sol>australian>hcp>single>tim>solid
Single: Aussie>HCP>Muon>Sol>Singletxguy>TimTurner>Omega>Solid4096/jimrtex

[jimrtex]

Here are the votes:
TimTurner: (not sure if the order or the numbers indicate the rank, but either way Sol's would win)
5. HCP
1. Sol
6. Solid4096/jimrtex
4. Scarlet
2. TimTurner
3. muon2
7. AustralianSwingVoter
8. Singletxguyforfun

Sol: Sol>muon2=Scarlet>AustralianSwingVoter>Solid>HCP=TimTurner
cvparty: scarlet>muon2>sol>australian>hcp>single>tim>solid
Single: Aussie>HCP>Muon>Sol>Singletxguy>TimTurner>Omega>Solid4096/jimrtex

The result is dependent on only having four judges.

[muon2]

Here are the votes:
TimTurner: (not sure if the order or the numbers indicate the rank, but either way Sol's would win)
5. HCP
1. Sol
6. Solid4096/jimrtex
4. Scarlet
2. TimTurner
3. muon2
7. AustralianSwingVoter
8. Singletxguyforfun

Sol: Sol>muon2=Scarlet>AustralianSwingVoter>Solid>HCP=TimTurner
cvparty: scarlet>muon2>sol>australian>hcp>single>tim>solid
Single: Aussie>HCP>Muon>Sol>Singletxguy>TimTurner>Omega>Solid4096/jimrtex

So if I understand correctly, applying condorcet to the four ballots Sol and muon2 tie. Then using IRV (which is not recommended when there are more candidates than voters) muon2 is eliminated on the first round with no 1st place votes along with all others with no first place votes. In the second round Scarlet and Aussie are eliminated with 1 vote each. That leaves only Sol. There is no reduction of one candidate at a time which is presumed by IRV, which is why it isn't used for so few voters.

[AustralianSwingVoter]

Colorado Non-Partisan plan.

My non-partisan redistricting plan for Colorado. Only 3 counties are split, and only Thornton (Adams) and Pueblo (Pueblo) cities are split.

District 1 D+24.23 - 74.1 - 24.3
District 2 D+09.33 - 61.4 - 36.9
District 3 R+08.20 - 46.6 - 51.6
District 4 R+13.19 - 41.8 - 56.7
District 5 R+11.41 - 41.7 - 56.8
District 6 D+03.52 - 55.1 - 43.3
District 7 D+02.72 - 54.8 - 43.3

[jimrtex]

Here are the votes:
TimTurner: (not sure if the order or the numbers indicate the rank, but either way Sol's would win)
5. HCP
1. Sol
6. Solid4096/jimrtex
4. Scarlet
2. TimTurner
3. muon2
7. AustralianSwingVoter
8. Singletxguyforfun

Sol: Sol>muon2=Scarlet>AustralianSwingVoter>Solid>HCP=TimTurner
cvparty: scarlet>muon2>sol>australian>hcp>single>tim>solid
Single: Aussie>HCP>Muon>Sol>Singletxguy>TimTurner>Omega>Solid4096/jimrtex

So if I understand correctly, applying condorcet to the four ballots Sol and muon2 tie. Then using IRV (which is not recommended when there are more candidates than voters) muon2 is eliminated on the first round with no 1st place votes along with all others with no first place votes. In the second round Scarlet and Aussie are eliminated with 1 vote each. That leaves only Sol. There is no reduction of one candidate at a time which is presumed by IRV, which is why it isn't used for so few voters.

This shows the results for Condorcet using several different methods. The only one that does not show a tie is Kemeny-Young, but it apparently includes a coin flip. Condorcet New Mexico

[Sol]

-snip-

Could you show a closeup on the Pueblo area too?

[muon2]

Here are the votes:
TimTurner: (not sure if the order or the numbers indicate the rank, but either way Sol's would win)
5. HCP
1. Sol
6. Solid4096/jimrtex
4. Scarlet
2. TimTurner
3. muon2
7. AustralianSwingVoter
8. Singletxguyforfun

Sol: Sol>muon2=Scarlet>AustralianSwingVoter>Solid>HCP=TimTurner
cvparty: scarlet>muon2>sol>australian>hcp>single>tim>solid
Single: Aussie>HCP>Muon>Sol>Singletxguy>TimTurner>Omega>Solid4096/jimrtex

So if I understand correctly, applying condorcet to the four ballots Sol and muon2 tie. Then using IRV (which is not recommended when there are more candidates than voters) muon2 is eliminated on the first round with no 1st place votes along with all others with no first place votes. In the second round Scarlet and Aussie are eliminated with 1 vote each. That leaves only Sol. There is no reduction of one candidate at a time which is presumed by IRV, which is why it isn't used for so few voters.

This shows the results for Condorcet using several different methods. The only one that does not show a tie is Kemeny-Young, but it apparently includes a coin flip. Condorcet New Mexico

Interesting that none of the ranked choice voting methods could definitively break the tie between Sol and muon2. I think that speaks to the need to hold off until all five votes are cast. If there aren't five active voters then the exercise should pause until 5 committed panelists are identified.

[muon2]

This is from my 2015 series built with the muon rules. I couldn't resist posting this from my hotel room in Golden CO.

There are only two county chops, both as required in the Denver metro, and they follow muni lines to the extent possible. The Denver UCC (Denver, Arapahoe, Jefferson, Adams, Douglas, Broomfield) has a population for 3.5 CDs. This plan uses the minimum number of CDs (4) to cover the Denver UCC and is one short of packing 3 CDs entirely within as the Jefferson CD includes tiny Clear Creek county.

Using political results available in 2010 (DRA didn't have the 2012/2016 PVI then) this plan had 3 Dem CDs, 3 Pub CDs and a tossup, and 1 each of the Dem and Pub CDs were competitive (PVI < 5.5).





Here are the population devs with the new PVIs.

CD 1: +818; D+24
CD 2: +1118; D+11
CD 3: -3552; R+7.6
CD 4: +876; R+13
CD 5: -469; R+14
CD 6: -729; D+5.4
CD 7: +1935; D+1.4

Since 2010 the competitive Pub seat has shifted right to be uncompetitive and the competitive Dem seat is just barely in that category. The suburban tossup seat has also almost shifted out to be a competitive Dem seat. This illustrates the changes that can happen over a decade from when a map would have been drawn, in this case showing increased urban-rural polarization.

[cvparty]

update I guess, I've been thinking about muon2's argument against IRV, and average deviation will instead be used in cases of ties. this means muon2's map is actually the winner for NM. not really a significant change since Sol's map is nearly identical

[Sol]

update I guess, I've been thinking about muon2's argument against IRV, and average deviation will instead be used in cases of ties. this means muon2's map is actually the winner for NM. not really a significant change since Sol's map is nearly identical

( )

[cvparty]

update I guess, I've been thinking about muon2's argument against IRV, and average deviation will instead be used in cases of ties. this means muon2's map is actually the winner for NM. not really a significant change since Sol's map is nearly identical

( )

sorry!!! your maps are awesome, you did win KY. for most states there's no truly single best map. most of the "winners" in this thing are just one best map, not the best map.