DRA stuff

[muon2]

Thanks, much nicer.

To give you a sense of the scoring and some changes you might consider, I'm just going to describe your CD 4. It is involved with 5 county chops: Eaton, Ingham, Isabella, Montcalm, and Saginaw. The Eaton and Saginaw chops each exceed 5% of a CD so they are macrochops. That means the disposition and boundaries of the individual subunits in those counties matter. This has the practical effect of treating them like the urban areas around Detroit and increases the erosity score that describes the shape of the district. Let me talk about each of the chops separately.

The Eaton and Ingham chops are what we call a "traveling chop". That means the same two districts chop more than one county. The rules can score it, but it's actually illegal under MI law, and can't be used unless required by federal law. There is an easy fix: put all of Ingham in CD 4, and shift the city of Eaton Rapids and the township immediately north to CD 4, too. Then shift Grand Ledge and the township around it out of CD 4. The traveling chop is gone and the total number of chopped counties drops to 4.

Isabella is a nice straight, simple chop. The chop intercepts the state highway connections from Mt Pleasant (county seat) to the county seats of both Midland and Gratiot, so it actually improves the erosity score. Well designed chops should do that.

Montcalm is a chop that creates two pieces. This tends to increase erosity, since each piece has a separate connection. A better design would chop out of CD-4 the 4 southernmost townships of Montcalm as well as the two western townships immediately north of those four. That arrangement serves to reduce the overall erosity.

Saginaw is another macrochop that will increase erosity by it's nature. There is also in CD 4 a precinct of a township west of Saginaw city that was chopped, and that increases the chop score by 1. Better is to leave that precinct out of CD 4 but use the township kitty-corner to the NW corner of Genesee instead.

See what you think of that, and if you want to make any of the changes I suggest. You may want to look at getting rid of those macrochopped counties, too.

[muon2]

Population deviations would be helpful. Even better would be to turn on city and town lines for the Detroit area. In this case an enlargement of the Grand Rapids area with town lines would be helpful to assess the map.

[muon2]

The SW part of your most recent MI plan is looking much better. Here's some analysis and thoughts on scoring for the northern part of that same plan.



I've redrawn just the three northern CDs (1 black, 2 green, 4 red). There are three chops between them:  CDs 1 and 4 chop Gladwin, and CDs 2 and 4 chop Grand Traverse and Kent. Note that this is another traveling chop between CDs 2 and 4 even though the chops are quite separated.

The little white lines show links that are cut between units in the plan. These add up to measure ersoity. Mostly they are between counties that are connected by state highways (see the map in the previously linked thread) or between the parts of chopped counties in different districts. Kent is macrochopped, so the links are local connections between subunits within Kent. There are 11 cut links between CDs 1 and 4, and 11 cut links between CDs 2 and 4, for a total erosity of 22.



I've redivided the same northern area to show how the scores could be reduced. CD 1 is whole counties and there is only one chop between CDs 2 and 4 in Montcalm. In addition the Grand Rapids UCC (Kent + Ottawa) is down to the minimum number of CDs. So this plan has an overall reduction of the CHOP score by 3.

There are now 3 cut links between CDs 1 and 2, 7 cut links between CDs 1 and 4, and 5 cut links between CD 2 and 4. This makes the erosity from those district boundaries equal to 15, so it reduces the EROSITY score by 7.

[muon2]

That is a fine boundary for CD 1. Excellent.

The chop in Kent county is extremely erose, however. It just looks like a gerrymander by protruding south the way it does. The boundary between CDs 2 and 3 is an 18. If you put Ottawa in CD 2 and Grand Rapids in CD you can reduce the erosity of their boundary to 12 and get rid of the chop in Wyoming.

Here's a map of the connections within Kent. Both yellow and blue connections apply between subunits in the county, but only yellow links are state highways that count between counties.



Also I note that Saginaw is a macroshop and you have a chopped township. If you move Gratiot to CD 8 and include more of Saginaw in CD 4 the macrochop is reduced to a simple chop and the erosity is lowered.

I looked at this this morning and i realized that your CD 2 could be made of whole counties by dropping all of Kent and adding Ottawa and Allegan. The extra population for CD 6 can come from a chop of Barry instead of Allegan. Happy 4th.

[muon2]

I don't want to tamper with the lines too much in Kent, or add Ottawa and Allengan, b/c the 2nd is designed to be a Swing district. (a solid D seat solely in such a region as Western Michigan seems impossible to me).  Though if that wasn't a concern I would redraw the lines.
Minor changes in Kent are fine, though. As long as all of Grand Rapids is in the seat, it won't be solid R. Perhaps add Grand Rapids Township and take out the rest of Wyoming?

The scoring rules are designed to punish political gerrymanders that have poor shapes. If you are going to gerrymander CD 2 I'd recommend putting Grandville, Wyoming and Kentwood in CD 3 and everything to their north in CD 2. That puts CD 2 at D+2 (55.2% Obama 08). The erosity along the border is 10, which is much better than the 18 in your most recent version. However the whole county version I suggested only has an erosity of 4.

You can also reduce erosity by combining the two fragments in the CD 6 chop of Allegan. If you don't like the combination in Allegan, a chop in Barry would score the same.

What're your thoughts about Saginaw? That large chop is costly in scoring, but shifting Gratiot to reduce the Saginaw chop helps a lot.

[muon2]

I don't want to tamper with the lines too much in Kent, or add Ottawa and Allengan, b/c the 2nd is designed to be a Swing district. (a solid D seat solely in such a region as Western Michigan seems impossible to me).  Though if that wasn't a concern I would redraw the lines.
Minor changes in Kent are fine, though. As long as all of Grand Rapids is in the seat, it won't be solid R. Perhaps add Grand Rapids Township and take out the rest of Wyoming?

The scoring rules are designed to punish political gerrymanders that have poor shapes. If you are going to gerrymander CD 2 I'd recommend putting Grandville, Wyoming and Kentwood in CD 3 and everything to their north in CD 2. That puts CD 2 at D+2 (55.2% Obama 08). The erosity along the border is 10, which is much better than the 18 in your most recent version. However the whole county version I suggested only has an erosity of 4.

You can also reduce erosity by combining the two fragments in the CD 6 chop of Allegan. If you don't like the combination in Allegan, a chop in Barry would score the same.

What're your thoughts about Saginaw? That large chop is costly in scoring, but shifting Gratiot to reduce the Saginaw chop helps a lot.

Does "everything north" mean all the county north or just areas to the west of Courtland Township?
Also I did play around with Saginaw County quite a bit, I wanted to avoid splitting Saginaw city itself. It took some time accomplishing that.
I hadn't considered shifting Gratiot. That definitely would have to be something I would have to consider.
What would the Saginaw chop look like if I did that?

I used a line across Kent county that goes through Lowell.

There are a couple ways of drawing the Saginaw chop with no chopped townships. To minimize erosity it probably helps to wrap the NW corner of Genesee that is in CD 8 with CD 8 in Saginaw.

[muon2]

This gets to the more interesting question mathematically. What is the maximum number of whole county CDs in any given state, assuming that the rest of the state can be divided into CDs as well and the CDs are within 0.5% of the quota? If there is more than one arrangement the one with the smallest difference in population between two CDs (the range) is best.

[muon2]

Non-partisan map of KS

1 (blue): This seat is the largest one in the state, and encompasses the vast majority of the state's population outside of the metro areas of Topeka, Kansas City, and Wichita. 29.7 O, 68.4 M; R+23.
2 (green): This district includes most of the outer reaches of the Kansas City metro not in the 3rd. It also includes all of the Topeka MSA as well as the rural Southeastern part of the state. 45.6 O, 52.5 M; R+7.
3 (purple): This seat is little altered from the current version. It barely changes at all, and thus retains its heavily urban territory. 48.7 O, 50 M; R+4.
4 (red): This seat centers on the five counties of the Wichita CSA (which it has the entirety of), and grabs part of Reno County, including all of Hutchinson. 40.6 O, 57.5 M; R+14.

Why would a non-partisan map chop Reno county, if populations don't have to be exact? There are plenty of combinations of other counties that don't chop it.

[muon2]

Non-partisan map of KS

1 (blue): This seat is the largest one in the state, and encompasses the vast majority of the state's population outside of the metro areas of Topeka, Kansas City, and Wichita. 29.7 O, 68.4 M; R+23.
2 (green): This district includes most of the outer reaches of the Kansas City metro not in the 3rd. It also includes all of the Topeka MSA as well as the rural Southeastern part of the state. 45.6 O, 52.5 M; R+7.
3 (purple): This seat is little altered from the current version. It barely changes at all, and thus retains its heavily urban territory. 48.7 O, 50 M; R+4.
4 (red): This seat centers on the five counties of the Wichita CSA (which it has the entirety of), and grabs part of Reno County, including all of Hutchinson. 40.6 O, 57.5 M; R+14.

Why would a non-partisan map chop Reno county, if populations don't have to be exact? There are plenty of combinations of other counties that don't chop it.

I liked the square ish shape that resulted. That was part of it. The 4th was being designed to be a more urban district (within reason) and once I drew in all of Metro Wichita I had a certain number of people left. I reached quota in a way that furthered my goals, I think.
What whole county district combinations regarding the 4th do you have in mind?

This keeps the 5-county Wichita area intact, eliminates the Reno chop, and reduces the overall erosity.



It does cut Wabaunsee out of CD 2, but that can be fixed at the cost of some erosity, but with better population equality.



In some detailed analysis a few of us did here a few years ago, we found the CSAs weren't the best measure of what made up a metro area for the purposes of keeping it together in a district. In general CSAs included too many rural counties, but those counties had enough commuters to count for Census purposes. We developed, and jimrtex codified, a definition for urban county clusters (UCCs) that are a sticky thread on the board. When we score plans one generally wants to cover a UCC with as few CDs as possible, and for large UCCs, pack as many CDs in them as possible.

[muon2]

What about this?

This looks significantly better. IN order to score it I wanted to make sure I could duplicate it. I get the following for the deviations: -383, +861, +2867, -3344. You probably have some tiny precincts in the wrong district. You can use the Find Unassigned Dists button to get rid of the remaining 2 population in unassigned. You can use the Check Contiguity button for each of the CDs to see if you left a bit of one district inside another. For pieces right at the border, I would increase the color opacity and perhaps change colors, to make it easier to spot little bits overlapping a county line.

Getting the populations right is used to get the inequality part of the score. The range is the difference from the largest to smallest population, and it is then put into this table to get a score.

The INEQUALITY score for a plan is found by taking the range for a plan and comparing it to the table below.

Range	Inequality
0-1	0
2-10	1
11-100	2
101-400	3
401-900	4
901-1600	5
1601-2400	6
2401-3200	7
3201-4000	8
4001-4800	9
4801-5600	10
5601-6300	11
6301-7000	12
7001-7700	13

The erosity is determined by counting the number of links between counties that are cut by boundaries between CDs. This is the connection map for KS:



For example, in your plan I count these links cut on boundaries coming to a total of 30:

CD 1 - CD 2: 9
CD 1 - CD 4: 11
CD 2 - CD 4: 6
CD 2 - CD 3: 4
Note that the path from Franklin to Miami counties, goes between their county seats on KS-68. That crosses between CD-2 and CD-3, but if Richland township is switched from CD 3 to CD 2 that removes that cut link. The population stays within the limits by my count.

[muon2]

Two other thoughts about KS.

Here's a neutral map I drew in 2014 with low erosity (27) and a range of just over 1700 (inequality 6).



I also found for your Dem gerry that if you take the whole counties of Wyandotte, Leavenworth, Atchison, Jefferson, Douglas, Franklin, Shawnee, Pottawatomi, Riley, and Geary the population is within 0.5% and the 2008 vote was Obama 53.2%, McCain 45.0% which is better than his Obama's national result. You can build a plan of whole counties for the remaining three CDs, too.

[muon2]

NM D gerrymander

1 (green): 46.9 White, 43 Hispanic; D+6
2 (blue): 43.9 White, 50 Hispanic; R+2
3 (purple): 45.1 White, 33.8 Hispanic, 17.9 Native; D+6

Why do you say it's D? It looks quite neutral to me. The state as a whole is D+4 so a 2D-1R split would be a fair division.

[muon2]

Is anyone else unable to see the 2008 election results by precinct in Oregon?

They were never part of DRA to my knowledge.

[muon2]

Is anyone else unable to see the 2008 election results by precinct in Oregon?

They were never part of DRA to my knowledge.

I'm not sure if 'precinct' is the right word, but for some reason the Election checkbox to see which areas voted for Obama and McCain won't work for me in Oregon.

That's because DRA loaded no election data for OR. Without the data, the button is disabled. If you notice OR in DRA only provides block groups, not VTDs as a option. In fact, if you go to the Census and try to pull up American Fact Finder data for VTDs in OR, there generally aren't any.

[muon2]

In 2011 I posted this map using DRA. It was redrawn from my submission to the Ohio Redistricting Contest where it was one of the top plans. For the contest the political measure was a composite of results from 2008 and 2010. District 11 was strong D, districts 1,5,7,13,14,15,16 were lean D, district 9 was even, districts 3,6,10 were lean R and districts 2,4,8,12 were strong R. Districts 1,3,5,7,9,14,15 were highly competitive with neither side exceeding +2.5%. In 2008 Obama won 11 of the 16 districts losing only the strong R districts plus lean R CD 6.



Now that the 2012/16 data is available I thought it would be interesting to compare those districtsand see how they have evolved this decade. The first number is the 2008-2010 composite used in the competition and the second is the DRA PVI.

CD 1: D+2.3 -> PVI D+4.5
CD 2: R+13.2 -> PVI R+17.9
CD 3: R+2.5 -> PVI R+2.7
CD 4: R+17.0 -> PVI R+22.8
CD 5: D+1.5 -> PVI R+0.3
CD 6: R+2.7 -> PVI R+12.9
CD 7: D+2.5 -> PVI D+2.6
CD 8: R+15.8 -> PVI R+17.0
CD 9: D+0.7 -> PVI R+2.3
CD 10: R+3.8 -> PVI R+3.8
CD 11: D+30.2 -> PVI D+31.8
CD 12: R+13.9 -> PVI R+13.1
CD 13: D+3.2 -> PVI D+2.7
CD 14: D+1.3 -> PVI R+2.4
CD 15: D+2.5 -> PVI D+9.0
CD 16: D+3.5 -> PVI R+2.9

There were a number of shifts here. CD 5 went from lean D to even. CD 6 went from lean R to strong R. CD 9 went from even to lean R. CD 14 went from lean D to lean R. CD 15 went from lean D to strong D. CD 16 went from lean D to lean R. So the plan went from 8 D, 7 R, 1 even to 5 D, 10 R, 1 even between 2010 and 2016.

[muon2]

An ungerrymandered Oregon legislative district map really should not give Republicans that good of a chance to take a majority.

Why? The Dems in OR are really concentrated in Portland so most fair geographic splits will make more overwhelmingly Dem districts (> D+20) than overwhelming Pub districts (> R+20). It doesn't serve those who want to eliminate gerrymandering to instead gerrymander to keep natural concentrations of one party from accumulating. You are viewing the map from a perspective that is very specific to this point in history.

[muon2]

6-district Oregon proposal

The nice thing about this map is that it creates three competitive seats, there are two strong D seats and one strong R seat, and three swing districts.

District 1 – 72.9% White, 14.1% Hispanic, 7.7% Asian, 1.5% Black, 0.6% Native
59.3% Obama, 38.6% McCain
D+8.39

District 2 – 81.3% White, 12.7% Hispanic, 2.1% Native, 1.1% Asian, 0.5% Black
55.9% McCain, 42.0% Obama
R+12.87

District 3 – 71.5% White, 10.4% Hispanic, 7.4% Asian, 6.0% Black, 0.8% Native
79.9% Obama, 18.0% McCain
D+31.58

District 4 – 85.5% White, 7.4% Hispanic, 1.9% Asian, 1.3% Native, 0.7% Black
54.5% Obama, 43.2% McCain
EVEN

District 5 – 80.3% White, 12.5% Hispanic, 3.2% Asian, 0.8% Black, 0.7% Native
52.4% Obama, 45.7% McCain
R+0.9

District 6 – 79.1% White, 13.4% Hispanic, 2.6% Asian, 1.3% Native, 0.8% Black
52.8% Obama, 44.9% McCain
R+0.87

I tried to eliminate the ridiculous gerrymander of District 5 by moving most of Tillamook County to the 1st district, and also made the 3rd entirely within Multnomah County.

Since there are 6 CDs, is this a potential plan for 2020? If so, are you using new population estimates or the old 2010 population numbers on DRA?

[muon2]

6-district Oregon proposal

The nice thing about this map is that it creates three competitive seats, there are two strong D seats and one strong R seat, and three swing districts.

District 1 – 72.9% White, 14.1% Hispanic, 7.7% Asian, 1.5% Black, 0.6% Native
59.3% Obama, 38.6% McCain
D+8.39

District 2 – 81.3% White, 12.7% Hispanic, 2.1% Native, 1.1% Asian, 0.5% Black
55.9% McCain, 42.0% Obama
R+12.87

District 3 – 71.5% White, 10.4% Hispanic, 7.4% Asian, 6.0% Black, 0.8% Native
79.9% Obama, 18.0% McCain
D+31.58

District 4 – 85.5% White, 7.4% Hispanic, 1.9% Asian, 1.3% Native, 0.7% Black
54.5% Obama, 43.2% McCain
EVEN

District 5 – 80.3% White, 12.5% Hispanic, 3.2% Asian, 0.8% Black, 0.7% Native
52.4% Obama, 45.7% McCain
R+0.9

District 6 – 79.1% White, 13.4% Hispanic, 2.6% Asian, 1.3% Native, 0.8% Black
52.8% Obama, 44.9% McCain
R+0.87

I tried to eliminate the ridiculous gerrymander of District 5 by moving most of Tillamook County to the 1st district, and also made the 3rd entirely within Multnomah County.

Since there are 6 CDs, is this a potential plan for 2020? If so, are you using new population estimates or the old 2010 population numbers on DRA?

Yeah, it's potential plan but I am using 2010 data

Because of the growth in the Portland area you won't need Tillamook in your CD 2 or Marion in your CD 4, especially it expands east instead. I've used DRA to get draw the map and get the PVI, but I used 2020 projections from the 2016 Census estimate data. This plan only splits one county other than Multnomah and keeps the deviation within 2000 from the quota.



I've matched the CD numbers up with the current districts, creating a new one for Portland.

CD 1: D+9
CD 2: R+11
CD 3: R+1
CD 4: R+1
CD 5: R+0
CD 6: D+32

There are 2 D, 1R, and 3 even CDs.

[muon2]

Utah new

District 1 (Blue): R+26.4
District 2 (Green): D+4.6
District 3 (Purple): R+25.5
District 4 (Red): R+33.5

I wanted to create a single district out of the rural areas of the state, which are currently divided between all 4, and gerrymandered to favor the Republicans. I also created a D-leaning district out of Salt Lake, Davis, and Summit counties.

This would also put Reps. Rob Bishop and Chris Stewart in a primary against each other for the new 1st district. Mia Love would be able to run in the 4th, and John Curtis in the 3rd. The 2nd wouldn't have an incumbent.

Is the Summit part of 4 connected to the rest of the CD? It's certainly contiguous, but it would be unfortunate if they were only contiguous over a mountain range with no passes there. Box Elder has a similar problem in that there's nothing but dirt trails to connect it with Tooele to the south.

[muon2]

Utah new

District 1 (Blue): R+26.4
District 2 (Green): D+4.6
District 3 (Purple): R+25.5
District 4 (Red): R+33.5

I wanted to create a single district out of the rural areas of the state, which are currently divided between all 4, and gerrymandered to favor the Republicans. I also created a D-leaning district out of Salt Lake, Davis, and Summit counties.

This would also put Reps. Rob Bishop and Chris Stewart in a primary against each other for the new 1st district. Mia Love would be able to run in the 4th, and John Curtis in the 3rd. The 2nd wouldn't have an incumbent.

Is the Summit part of 4 connected to the rest of the CD? It's certainly contiguous, but it would be unfortunate if they were only contiguous over a mountain range with no passes there. Box Elder has a similar problem in that there's nothing but dirt trails to connect it with Tooele to the south.

Yeah, there's interstate 80.

I mean it looks like you can't drive to the I-80 part of CD 4 in Summit county without driving through CD 2. Some states require that it is convenient to get to all parts of a district without going through other districts. It's a good neutral redistricting principal that can cut down on certain types of gerrymanders.

[muon2]

I mean it looks like you can't drive to the I-80 part of CD 4 in Summit county without driving through CD 2. Some states require that it is convenient to get to all parts of a district without going through other districts. It's a good neutral redistricting principal that can cut down on certain types of gerrymanders.

Oh, for some reason I thought you meant CD2. Well, you can take State Road 32, which barely goes through CD2 for some parts.

We've had lots of thread debates about whether a brief cut into another district disqualifies. We've looked at formulas to define how much is too much in the other district. In the end, my conclusion is that if you split a county the pieces should be reachable from their respective districts without cutting any length into another district.

[muon2]

Here's a new Tennessee

District 1 (Blue): R+28.2, current R+28
District 2 (Green): R+19.1, current R+20
District 3 (Purple): R+19.3, current R+18
District 4 (Red): R+24.0, current R+20
District 5 (Yellow): D+4.5, current D+7
District 6 (Turquoise): R+20.5, current R+24
District 7 (Gray): R+20.5, current R+20
District 8 (Light Purple): R+16.2, current R+19
District 9 (Light Blue): D+22.4, current D+28

Out of curiosity, why chop both Hamblen and Hawkins between CD 1 and 2? It's always possible to replace it with one chop in one of those two counties. The same question applies to CD 4 and 6, between 4 and 7, and between 7 and 8.

[muon2]

No Counties Split; Max Deviation 1000; West Iowa Intact; Des Moines Metro Intact; Iowa Districts



Blue: D+1.7
Green: R+14.36
Red: D+4.46
Yellow: R+3.02

I could have made the Red and Yellow Districts much neater looking by splitting even a single County.

IA has 4 districts, 99 counties, and no county larger than a CD. Mathematically one expects that a whole county division based on those stats should have a range of less than 100 between the largest and smallest CD. The enacted IA plan was within that limit with deviations of -41, +35, +23, and -18. It would be hard to have the court support a plan that had a significantly larger range unless it was decidedly more compact.

[muon2]

Last decade they used 2008 estimates in the map, but never projected them to 2010. DRA also only used the county estimates, so they spread the population changes equally to all precincts in each county. If they do the same this decade, I expect 2018 estimates appearing in the summer of 2019, but no 2020 projections.

[muon2]

I thought this was a DRA stuff thread.