Forum Redistricting Commission

[muon2]

Let me propose the following. A commission of five or seven members is determined by lot from among those interested with no party having a majority. The commission agrees to the criteria and software (presumably DRA) and is in charge of making sure that the criteria are followed. They can also be arbiters of things like VRA compliance. I'd be happy to act as a consultant on measurable criteria, based on the various rubrics we've tested here in the past.

Once the criteria are set for a state then the public (including the commission) can propose maps for the state for a fixed period of time. Proposed maps are posted here so that the public can view all maps, and propose new maps that improve upon existing proposals. At the end of the time period the commission votes from those maps that best meet the criteria. In this way the map is crowdsourced under the auspices of the commission, and everyone gets some say in the process.

As to the state, I would suggest VA since its congressional plan was just struck down by the court, pending appeal.

[muon2]

Here's who I have as applicants for the commission with affiliation based on avatar. You can add or remove your name as you wish. I propose that applications be accepted until Friday 11/21/14 11:59 pm EST.

Sol (I)
JerryArkansas (R)
SLCValleyMan (G)
Dixie (R)
Frodo (D)?
Miles (I)
Del Tachi (R)
traininthedistance (D)??
X (G)

[muon2]

Ugh, I wish my DRA still worked.

But you can still be selected for the commission.

[muon2]

Also what preferences does everyone have for states? The biggest question to figure out is the number of districts we want to work with, I think. Which categories in the following tier list are acceptable to y'all? At what point, if any, will the exercise become too big or too small?

Big Four: CA/TX/FL/NY
Large States: IL/PA/OH/MI/GA/NC
Medium States: NJ/VA/WA/AZ/IN/TN/MA
Small States: MD/WI/MO/MN/AL/SC/CO
Tiny States: KY/LA/CT/OK/OR
Simple States: the rest (four districts or less)

I still think VA is most interesting since it is currently under scrutiny by the court. If the VRA makes the issue too complex in VA, I'd stick to states that are medium or smaller for an initial run.

[muon2]

Some would like to start with a smaller state, so let me add another specific vote to take. I'll pick three smaller states that could also make interesting candidates that had legislative gerrymandering. If simpler is better to start, it's helpful if they don't involve the VRA and do have lots of internal political units to guide map making.

KY: 6 CDs, no VRA issues, lots of counties, bipartisan compromise for incumbents.

WI: 8 CDs, no VRA issues, townships to guide county chops, Pub gerrymander.

IN: 9 CDs, no VRA issues, townships to guide county chops, Pub gerrymander.

VA: 11 CDs, VRA issue, lots of counties and independent cities, Pub gerrymander and court challenge.

Since the non-commissioners will also be participating with maps, I think everyone should vote for the state. I like the idea of approval voting, so vote for any number of the states in the list.

[muon2]

There is currently a tie between VA and WI for the first state for the commission. Please vote for just one of those two to break the tie. Voting will close at 11:59 EST tonight.

The following posters have applied for the commission (I think). Some posters voted for the state, but did not clearly indicate if they wanted to be considered for the commission. If that included you and you meant to be on the list, please indicate some time today. If you didn't mean to be on the list, also indicate that today. Tomorrow I'll make a random draw of five commissioners and two alternates such that no party gets a majority.

JerryArkansas (R)
Dixie (R)
Del Tachi (R)

Averroes Nix (D)
Morgieb (D)
Gass (D)
Bacon King (D)
X (D)

SLCValleyMan (G)
Fuzzybigfoot (G)

Electionsguy (L)

Sol (I)
Miles (I)

[muon2]

The commissioners are

Morgieb
Miles
Fuzzybigfoot
Dixie
Del Tachi

and the alternates are

ElectionsGuy
X

I still need some tie breaking votes for the state (WI or VA).

[muon2]

Great to see so many lobsters. They'll get to submit their plans to the commission, too.

Once the state is selected (yes it's still tied by my count), the commission will have to vote on the criteria to produce a map.

My thoughts on commission votes is that it takes 3 votes to pass the commission. The alternates may vote, but their votes will only count if one or two commissioners fails to vote on a matter before the commission. In the spirit of crowdsourcing the final product, other posters may urge the commission to vote in a particular way during the discussion/voting period. Commissioners can vote at any time during the discussion/voting period and the vote can be changed, with only the last vote counted.

[muon2]

Fuzzy approves.

Is this your vote? If so, are you voting on all three items before the commission?

[muon2]

Here's my initial Virginia map.







District 1 (Blue)

• Deviation: 290
• Election 2008: John McCain - 54.4%, Barack Obama - 44.7%
• Racial Breakdown: 73% White, 15.3% Black, 6.2% Hispanic

District 2 (Green)

• Deviation: -422
• Election 2008: Barack Obama - 49.8%, John McCain - 49.4%
• Racial Breakdown: 62.6% White, 23.3% Black, 5.9% Hispanic

District 3 (Purple)

• Deviation: -30
• Election 2008: Barack Obama - 64%, John McCain - 35.2%
• Racial Breakdown: 48.8% White, 35.2% Black, 5.9% Hispanic

District 4 (Red)

• Deviation: 83
• Election 2008: Barack Obama - 67.6%, John McCain - 31.7%
• Racial Breakdown: 51.8% Black, 39.3% White, 5.2% Hispanic

District 5 (Yellow)

• Deviation: -903
• Election 2008: John McCain - 51.2%, Barack Obama - 47.8%
• Racial Breakdown: 71.7% White, 21.5% Black

District 6 (Teal)

• Deviation: 586
• Election 2008: John McCain - 56.4%, Barack Obama - 42.5%
• Racial Breakdown: 83.9% White, 7.9% Black, 5.1% Hispanic

District 7 (Silver)

• Deviation: 352
• Election 2008: John McCain - 57.9%, Barack Obama - 41.3%
• Racial Breakdown: 76.2% White, 13.1% Black

District 8 (Slate Blue)

• Deviation: 384
• Election 2008: Barack Obama - 68.2%, John McCain - 30.9%
• Racial Breakdown: 52.8% White, 12.9% Black, 18.2% Hispanic 19.7%, 11.5% Asian

District 9 (Light Blue)

• Deviation: 501
• Election 2008: John McCain - 59.5%, Barack Obama - 39.1%
• Racial Breakdown: 91.8% White

District 10 (Pink)

• Deviation: -219
• Election 2008: Barack Obama - 57.4%, John McCain - 41.9%
• Racial Breakdown: 61.4% White, 6.2% Black, 11.6% Hispanic, 17.3% Asian

District 11 (Light Green)

• Deviation: -624
• Election 2008: Barack Obama - 58%, John McCain - 41.3%
• Racial Breakdown: 51.3% White, 14.8% Black, 18.2% Hispanic, 12% Asian

Things I like about my map:
-District 6 is pretty much just Shenandoah Valley + Roanoke, making a very clean CoI
-Clean Majority-Minority district (VA-03), along with a black majority district (VA-04)
-Very few splits around Virginia Beach/Norfolk/Newport area. Most maps I've seen cut this area to shreds.
-5 Solid Democratic districts, 4 Solid Republican districts, 2 competitive Districts (VA-02 and VA-05)

Dislikes:
-I'd like to keep Richmond whole, but doing so would force multiple splits around Chesapeake Bay in order to form a majority-black district.
-More county splits than probably necessary.[/list]

Nice work, particularly describing your pros and cons. Some of those, like county chops, should become quantified once the scoring system is approved. Normally a commission wouldn't take submissions until all the rules are approved, but in a crowdsourcing environment, submissions will happen as soon as web links are up. However, if item 1 is adopted, you'll need to resubmit your detail maps with City/Town lines turned on to be considered by the commission. Chops of municipalities within counties will be proposed as part of the measures, so they need to be assessed.

[muon2]

Muon2, are you going to use the state highway chop count as a proxy to measure erosity? Do highway chops within chopped counties count for this category?  Is there going to be some pareto optimality metric?

The Pareto optimality provision is embodied in item 3 above. It accepts all maps that are on the Pareto frontier defined by chop and erosity scores. Inequality is used to reduce the number of maps that would be at the same point on the frontier.

I have spent a bit of time analyzing my earlier work on erosity. At the county level, I will propose the cut set of state highways that we have looked at before. I have some suggested refinements for mega-chopped counties such as Fairfax so that erosity is reduced in heavily populated areas where state highways are no longer the best metric. There will also be some discussion about what constitutes connectivity in the Hampton Roads area and I'm working on a primer to facilitate that. There may also be some amendments proposed to deal with independent cities that were forced by statute to take over some of their state highways.

[muon2]

I approve of the three items; and commend the excellent work on the part of angrygreatness!  Btw, Electionsguy should probably be added to the commission since I believe Dixie (a Republican commissioner) said he was leaving the forum permanently last night.

I see Dixie has just posted a departure thread on FC, so that would seem to confirm your statement. Unless I hear otherwise in the next day, I will replace Dixie. That moves ElectionsGuy onto the commission and you (X) to first alternate. I've drawn SLCValleyMan as the new second alternate.

[muon2]

I just want to remind the commissioners (and alternates) to vote on the three items before them. I'll continue with detailed implementation items after this basic outline is agreed to.

[muon2]

My submission for VA. My CD3 is 50.6% BVAP

How did you get the Warner-Gillespie numbers for the CDs with chopped counties/cities?

[muon2]

^ The Virginia Public Access Project has precinct-level results.

Very nice. Of course it helps that VA doesn't have too many precincts. It looks like my county of less than 1 million has more precincts than the whole state of VA.

[muon2]

To illustrate the above items 4 and 5, I will apply them to angryGreatness-A (I'll use letters to designate different plans from the same poster).

District 1 (Blue)

• Deviation: 290
• Election 2008: John McCain - 54.4%, Barack Obama - 44.7%
• Racial Breakdown: 73% White, 15.3% Black, 6.2% Hispanic

District 2 (Green)

• Deviation: -422
• Election 2008: Barack Obama - 49.8%, John McCain - 49.4%
• Racial Breakdown: 62.6% White, 23.3% Black, 5.9% Hispanic

District 3 (Purple)

• Deviation: -30
• Election 2008: Barack Obama - 64%, John McCain - 35.2%
• Racial Breakdown: 48.8% White, 35.2% Black, 5.9% Hispanic

District 4 (Red)

• Deviation: 83
• Election 2008: Barack Obama - 67.6%, John McCain - 31.7%
• Racial Breakdown: 51.8% Black, 39.3% White, 5.2% Hispanic

District 5 (Yellow)

• Deviation: -903
• Election 2008: John McCain - 51.2%, Barack Obama - 47.8%
• Racial Breakdown: 71.7% White, 21.5% Black

District 6 (Teal)

• Deviation: 586
• Election 2008: John McCain - 56.4%, Barack Obama - 42.5%
• Racial Breakdown: 83.9% White, 7.9% Black, 5.1% Hispanic

District 7 (Silver)

• Deviation: 352
• Election 2008: John McCain - 57.9%, Barack Obama - 41.3%
• Racial Breakdown: 76.2% White, 13.1% Black

District 8 (Slate Blue)

• Deviation: 384
• Election 2008: Barack Obama - 68.2%, John McCain - 30.9%
• Racial Breakdown: 52.8% White, 12.9% Black, 18.2% Hispanic 19.7%, 11.5% Asian

District 9 (Light Blue)

• Deviation: 501
• Election 2008: John McCain - 59.5%, Barack Obama - 39.1%
• Racial Breakdown: 91.8% White

District 10 (Pink)

• Deviation: -219
• Election 2008: Barack Obama - 57.4%, John McCain - 41.9%
• Racial Breakdown: 61.4% White, 6.2% Black, 11.6% Hispanic, 17.3% Asian

District 11 (Light Green)

• Deviation: -624
• Election 2008: Barack Obama - 58%, John McCain - 41.3%
• Racial Breakdown: 51.3% White, 14.8% Black, 18.2% Hispanic, 12% Asian

CD-01: PVI-08=-8.6; uncompetitive R
CD-02: PVI-08=-3.5; competitive R
CD-03: PVI-08=+10.8; uncompetitive D
CD-04: PVI-08=+14.4; uncompetitive D
CD-05: PVI-08=-5.4; competitive R
CD-06: PVI-08=-10.7; uncompetitive R
CD-07: PVI-08=-12.1; uncompetitive R
CD-08: PVI-08=+15.1; uncompetitive D
CD-09: PVI-08=-14.0; uncompetitive R
CD-10: PVI-08=+4.1; competitive D
CD-11: PVI-08=+4.7; competitive D

SKEW: +5-6 = -1; S score 1 (R)
POLARIZATION: 4+2*7 = 18; P score 18

[muon2]

For comparison, here are the political measures for Miles-A. For shorthand I use D and R for uncompetitive districts, d and r for competitive districts, and e for highly competitive districts. In order they are r,e,D,r,r,R,R,D,R,r,d or 2D, 1d, 1e, 4r, 3R. That gives S = 4 (R), P = 15.

[muon2]

"A competitive district has a PVI of 2 through 5 (-0.054 to -0.015 and +0.015 to 0.054) and statistically such districts have an 3 out of 4 chance of being held by the favored party."

That seems like a very wide range to me, with 1.5 to 2.5 quite competitive, and 4-5 not much, although I suppose an examination of results in the past would be instructive. In my brain, once you hit 4, a seat is pretty safe absent unusual developments, or a significant trend going on. Is there an advantage in bifurcating this category, and adjusting the point count some?

There was a 538 analysis some time ago and it matched the Cook use of PVI 5 or less as a "swing" seat. Statistically there's not as much difference between PVI 2 and PVI 4 as you might think. I could've made a fancy linear fit to the data so that the scale slides for all PVI's, but I couldn't find that it really adds anything. There's a tendency to micro-manage the statistics like when 538 says that a seat has moved from 66% R to 70% R. Mathematically it's correct, but given the statistical uncertainty of the input polls, it's not a meaningful change. If I can find the data set again I'll supply a link.

OTOH, there's a more meaningful issue inside the measure for SKEW that the commission may want to consider. If you look at the two maps I scored, you'll see that Miles-A is a more competitive plan than angryGreatness-A. However, that came at the expense of a Pub skew to Miles map. If the commission wants to reward competitiveness more, I would adjust SKEW so that uncompetitive districts only count twice as much as competitive districts. That would have no effect on angryGreatness-A since there are equal numbers of competitive d and r districts, but it would reduce the relative SKEW of Miles-A in comparison.

[muon2]

To aid people who are already working on maps and are looking at chopping Richmond to comply with the VRA, here is a map of the city divided into five areas based on the generally accepted groupings of neighborhoods. The boundaries are adjusted to conform with the voting districts in DRA.



Blue - Downtown; pop 11,816; BVAP 30.7%
Green - East End; pop 27,084; BVAP 71.1%
Purple - Northside; pop 32,018; BVAP 69.2%
Red - West End; pop 49,504; BVAP 14.2%
Yellow - Southside; pop 83,792; BVAP 54.4%

[muon2]

A gentle reminder to the commission that Items 4 and 5 are awaiting action and discussion. In particular I noted a possible amendment to Item 4 that may be worth discussion.

[muon2]

Is there a way to get county splits in DRA?

I was going to increase the number of CDs, then color the extra "CDs" for the county parts.  This should give me demographic and political data for the split county parts.

But when I increment the number of districts, all the existing districts disappear.

Unfortunately no. The best I've found is to either reserve many additional districts and keep track of the real quota on my own spreadsheet, or to make a number of separate DRA files after the fact to highlight the chops.

There is one shortcut to all this that I've used quite effectively. Once the map is complete, I double check with the Find Unassigned Dists tool. After everything is assigned I can unassign any county fragment and look at the demographics of both the fragment and the remaining district. It can be time consuming for a lot of fragments, but it avoids clearing the original map.

[muon2]

Since no commissioners have weighed in on the items before the commission, I'm not sure how to proceed. The process will drag out, and there will probably be a lack of interest if things don't keep moving. Commissioners morgieb, Miles, Del Tachi, and ElectionsGuy have all posted to the Atlas since Items 4 and 5 were posted, but not on those items. Only Fuzzybigfoot has been off since 11/26, and that's why there are two alternates: X and SLCValleyMan. Any feedback into how to get this on track is welcomed.

[muon2]

Since no commissioners have weighed in on the items before the commission, I'm not sure how to proceed. The process will drag out, and there will probably be a lack of interest if things don't keep moving. Commissioners morgieb, Miles, Del Tachi, and ElectionsGuy have all posted to the Atlas since Items 4 and 5 were posted, but not on those items. Only Fuzzybigfoot has been off since 11/26, and that's why there are two alternates: X and SLCValleyMan. Any feedback into how to get this on track is welcomed.

If the Virginia commission were operating under the terms of the Florida constitution, items 4 and 5 would be outlawed since they are politically-based.

The current litigation in Florida was whether or not the legislature was indirectly influenced to make political decisions.   The problem in Florida is that the legislature is a political body, and you would hope that they would be susceptible to political arguments.

In Britain, where the boundary commissions are independent, public representations by political parties are quite acceptable and ordinary, but they have to be cast in terms that the commission can recognize, such as communities of interest.

I would suggest that items 4 and 5 not be used directly by the commission, but that we (the forum community) use them as metrics for the process as a whole.

I included them as measures since political data is explicitly required by the AZ commission and were part of the proposal for an OH commission. Good government groups I've spoken with are divided as to the need to use or forbid the use of the political measures of a plan. Personally, I think that clever mappers will find a way to gain a political advantage even with extensive neutral rules (see MI). Political measures provide one means to determine if the rules have been gamed.

However, it is because of the political nature of this data that I placed SKEW and POLARIZATION as after-the-fact measures in Item 3. They can not be used to exclude a plan, but can be used to guide the commission's final selection. I also believe that commissioners will have inherent biases towards certain plans and the political data simply provides a check on those inherent biases.

[muon2]

Since no commissioners have weighed in on the items before the commission, I'm not sure how to proceed. The process will drag out, and there will probably be a lack of interest if things don't keep moving. Commissioners morgieb, Miles, Del Tachi, and ElectionsGuy have all posted to the Atlas since Items 4 and 5 were posted, but not on those items. Only Fuzzybigfoot has been off since 11/26, and that's why there are two alternates: X and SLCValleyMan. Any feedback into how to get this on track is welcomed.

If the Virginia commission were operating under the terms of the Florida constitution, items 4 and 5 would be outlawed since they are politically-based.

The current litigation in Florida was whether or not the legislature was indirectly influenced to make political decisions.   The problem in Florida is that the legislature is a political body, and you would hope that they would be susceptible to political arguments.

In Britain, where the boundary commissions are independent, public representations by political parties are quite acceptable and ordinary, but they have to be cast in terms that the commission can recognize, such as communities of interest.

I would suggest that items 4 and 5 not be used directly by the commission, but that we (the forum community) use them as metrics for the process as a whole.

I included them as measures since political data is explicitly required by the AZ commission and were part of the proposal for an OH commission. Good government groups I've spoken with are divided as to the need to use or forbid the use of the political measures of a plan. Personally, I think that clever mappers will find a way to gain a political advantage even with extensive neutral rules (see MI). Political measures provide one means to determine if the rules have been gamed.

However, it is because of the political nature of this data that I placed SKEW and POLARIZATION as after-the-fact measures in Item 3. They can not be used to exclude a plan, but can be used to guide the commission's final selection. I also believe that commissioners will have inherent biases towards certain plans and the political data simply provides a check on those inherent biases.

In Arizona, the use of the rule may have led to skewing of the results.  With so much of the population in a few counties, it is unlikely that but for that measure the competitive seats would tend to be underpopulated.

And it may be contrary to good representation.   Putting Cuyahoga and Holmes counties into a single district does not lead to good representative, even if it somehow led to a competitive race.  The representative can not be representative of the district, nor effectively represent the district since its interests are so disparate.  The initial redistricting initiative in Ohio may have failed because of the maps  that were produced showing you can get a long way from Cleveland without splitting counties.

AZ suffered from the mandate of competitiveness, not its inclusion. Since the commission was compelled to maximize the number of competitive districts, they were sensitive to a data set that was prone to skew their resulting districts. The maps produced by the opponents of the 2005 OH initiative also used the maximization of competitive districts to get strange results. The competing initiatives in 2010 which did not reach the ballot used political data as part of a mix of factors and did not prioritize it in a way that was likely to warp the final plan.

[muon2]

As for Arizona, while the map was basically a pretty fair map.  Four safe Republican seats, two safe Democratic seats, one competitive seat that has a small Democratic tilt, and two competitive seats with a slight Republican tilt seems like about what you'd expect from a fair Arizona map.  It just happened that a fair map benefited Democrats more in Arizona just as it would have benefited Republicans more had their been a truly independent redistricting commission in Illinois. 

The potential skew in AZ is due to the use of the 2008 data in the PVI's. The Pubs appear to have overperformed there due to McCain as the nominee. Until the data rolls off after the 2016 cycle, it's hard to call swing districts there very accurately. IL has the same problem, and it will be interesting to see the next set of PVI's when non-Obama data replaces 2008.