Dave's Redistricting App

[muon2]

I spent a couple of hours using the tool in OH where I've spent a good deal of time looking at maps during the last two months. The tool was interesting but it was cumbersome compared to real GIS software. I couldn't grab groups of blocks easily, and I couldn't find pieces that were inadvertently missed. The tool also assumes that within a county each block will change as the county does, so it really can only approximate what has happened this decade within some of the urban counties.

[muon2]

I encountered the same problem with polygons in Franklin County, OH where Columbus has lots of unincorporated pocket. It was so bad there that the actual boundaries of Columbus and most of its suburbs were obscured.

The population in the puce-colored district is along the coast south of Miami, with little south of Homestead, plus the population on the Keys.  It should simply be thought of the district south of the purple district.  I probably could have done a better job of following city boundaries but the application has problems with polygons with enclaves.  So you could keep all of Miami in a district.

[muon2]

4-district Iowa:



CD1 (blue) - 58.5 - 40.4 Obama
CD2 (green) - 58.0 - 40.8 Obama
CD3 (purple) - 55.0 - 43.6 Obama
CD4 (red) - 53.7 - 45.2 McCain

Basically, I ended up preserving two Democratic-leaning districts, a Republican district, and a Dem-leaning swing district. I have to say, Iowa is ridiculously easy to redistrict -- nice square lines, decentralized population, homogenous population. I don't know how strict their requirements are, but I got all the districts within the same population by a thousand or so except for the 3rd, which is 5,000 under. I also didn't take into account where any of the Congresspeople live -- Dave Loebsack lives in Linn County, apparently, so he'd have to move one county south to stay in CD2, and Tom Latham is in Story County, which put him in the 3rd with Boswell.

Don't worry about where IA congressmen live. IA doesn't use incumbent residences as a factor. Because IA is easy, you should try to get the districts to within about 100 people. Last time the deviations were +37, -24, +40, +40, and -94 from the ideal population, I assume that the mapmakers will try for the same goal next time.

[muon2]

Don't worry about where IA congressmen live. IA doesn't use incumbent residences as a factor. Because IA is easy, you should try to get the districts to within about 100 people. Last time the deviations were +37, -24, +40, +40, and -94 from the ideal population, I assume that the mapmakers will try for the same goal next time.

The legislature actually rejected the first plan because its deviation was too large (I think around 150).  I always figured they just didn't like what the computer drew, but had to have a rationale for rejecting the map.

Apparently the idea that a computer drew the map is somewhat of a myth. I spoke to someone who was involved, and learned that the IA maps were actually drawn by hand on paper. The computer was used to check variances for a court challenge.

[muon2]

Arizona with 8 districts.  Maybe if Arizona had an independent redistricting commission they could get something like this rather than the mess that they have.

Part of the problem is Arizona's policy on Native Americans with regards to redistricting.  I doubt even an independent commission would put the Hopi and Navajo tribes in the same district as you've done.

I've talked with one of the experts who worked on the 2000 remap. Padfoot is quite correct in his assessment of the issue.

[muon2]

I have not been able to create jpg files for my maps. I save the state map from the app, and I can find it on my drive. However, the map tp jpg converter on the site can't open the file. Has anyone else encountered this, and if so what fix is there?

[muon2]

I have not been able to create jpg files for my maps. I save the state map from the app, and I can find it on my drive. However, the map tp jpg converter on the site can't open the file. Has anyone else encountered this, and if so what fix is there?

That's another confusing thing about this app -- you use the save on the left-hand size to save the actual map, but the save on the right side to save pictures of the map that the converter uses.

After three tries withe left control for the state, I was able to get something that would open and convert to jpg. I made a zoomed area, saved the area, and I still don't get something I can open and convert, at best I get the state map back.

[muon2]

I challenge people to do maps, for fun, with Canadian (100,000) or French (125,000)-sized constituencies.

Iowa with 29 districts, with average population of 100,908, maximum deviation 4.0%, average deviation of 1.9%.

6 counties have a population in excess of one district and one or more whole districts was formed within: Polk County (Des Moines) 3.712; Linn (Cedar Rapids) 1.900; Scott (Davenport-Bettendorf) 1.572; Black Hawk (Waterloo-Cedar Falls) 1.269; Johnson (Iowa City) 1.100; Woodbury (Sioux City) 1.029.  Woodbury was not split, since it is only slightly over the limit.

Jackson County was also split.  Dubuque County (Dubuque) was short by about 11.7%, and adding the entirety of a neighboring county would create an excess of at least 6.5%.  The area of Jackson County placed with Dubuque County is largely along the Mississippi River south of Dubuque.

Des Moines, unlike most large city has not been fused with its township so Polk County was divided by townships.  Western Des Moines is Des Moines township with an almost ideal population of 100.4% of the target district population.  Most of the rest of the city is included in Lee township (east) and Bloomfield township (south).  This eastern district also includes Pleasant Hill and Carlisle.  The western suburban district includes large parts of West Des Moines, Windsor Heights, Clive, Urbandale, Grimes, and Johnston.  The northern suburban district, which includes Boone County, contains Ankeny, Saylorville, Altoona, and Bondurant.

Cedar Rapids is the only city that was split.  With a population of 119.6% of the target, area in the far southern part of the city were trimmed off and placed with the rest of Linn County and northern Johnson County.  Johnson County only has a slight overage, and the boundary is north of Iowa City and its suburbs.

Davenport is almost the ideal size of a district (97.5%), leaving the rest of Scott County, including Bettendorf to be paired with Muscatine County.

Waterloo and Cedar Falls together have a population equivalent to 104.0% of the target population, leaving the relatively small excess population of Black Hawk County to be placed in a district with 3 other counties.

A different question with IA has to do with their reapportionment. Assuming that they lose one congressional seat, what happens to the nesting of Senate and House seats within each CD? Currently the 5 CDs are divided into 10 Senate districts and each Senate district into 2 House districts. 50 doesn't divide into 4 so either IA changes the nesting rules or the number of State Senators.

[muon2]

10 districts for Mass on 2000 population.   Boston, 3 inner suburban, 3 outer suburban-exurban, Worcester-Springfield, Western Mass-Springfield, and Fall River-New Bedford-Cape Cod-Islands

A New England Republican might even win the red district.

[muon2]

I'm surprised nobody's done North Carolina yet. Here it is with a theoretical 14th district (which may or may not happen this time). You may want to open it in a new window to see it better.



As you can see, I've actually de-gerrymandered it to some extent. Here's the district breakdown:

NC-01 (brown, GK Butterfield - D) - I didn't bother trying to make this majority-black; it's a coalition district. I'm sure you could finagle it to 51% black if needed; currently it's about 49% or so white. Either way, easy Dem district to hold.
NC-02 (very light purple, Bob Etheridge - D) - I'm not sure how strong this district is -- it takes in half of Wake County (Raleigh) and Republican Johnston County. Probably a tossup district now.
NC-03 (pink, Walter Jones - R) - Extends the entire coastline, but is at least more contiguous now. Safe Republican.
NC-04 (green, David Price - D) - Still safe Dem, although I changed its boundaries to include Republican Alamance County, which is overwhelmed by Chapel Hill and Durham.
NC-05 (dark purple, Virginia Foxx - R) - Safe Republican district. Didn't change much.
NC-06 (light purple, Howard Coble - R) - I excised all of Greensboro, so Coble would have to move to the suburbs (or retire, he's 79), but it's safe Republican.
NC-07 (red, Mike McIntyre - D) - This was annoying, because there's not much Democratic territory in the southernmost point of North Carolina. I kept most of his district but gave Brunswick County to Jones. Still leans Republican.
NC-08 (light blue, Larry Kissell - D) - Shifted eastwards, picking up more of Fayetteville and some Republican areas of McIntyre's district. Probably a little more Republican now.
NC-09 (yellow, Sue Myrick - R) - She'll have to move to the suburbs too, but she gets a safe district still.
NC-10 (dark green, Patrick McHenry - R) - The district shrunk but is still safe Republican.
NC-11 (dark blue, Heath Shuler - D) - Also shrunk a bit, and is maybe a little more Democratic now that it doesn't have a few of the district's easternmost counties.
NC-12 (teal, Mel Watt - D) - Controversy! I dismantled his plurality-white district that snakes from Charlotte to Winston-Salem and gave him a safe, albeit majority-white, metro Charlotte district. I'm sure it probably violates VRA, but oh well.
NC-13 (magenta, Brad Miller - D) - This one shrunk as well, but remains centered in Raleigh.
NC-14 (grey, open) - A new Democratic seat that consists of Greensboro and Winston-Salem. Thanks, Mel!

Since the SCOTUS just ruled to continue requiring preclearance for states like NC, the lack of any majority black district would kill this plan. There will be a real challenge in NC to meet the VRA but not create a gerrymander which will fail on other grounds.

[muon2]

I had some time to play around with the tool this afternoon and looked at the map of MA using the 2008 data. I divided no town except for Boston as this is the primary goal of any MA map. All the districts are within 300 persons of the ideal size which is less than 0.1% maximum variation. I attempted to balance two secondary goals: to keep districts somewhat compact and to keep county fragments to a minimum.



Some items of interest compared to the current map which loses CD 10.

CD 1 picks up Springfield and loses the northern part.

CD 2 shifts east to pick up Worcester and Franklin County.

CD 3 becomes more compact and links the Fitchburg area to Framingham and Dedham.

CD 4 shifts south and east to take in the Cape and Islands from current CD 10.

CD 5 remains centered on Lowell and Lawrence, but dips south to the Woburn area.

CD 6 extends south all the way into the northern part of Boston including Beacon Hill.

CD 7 shifts south to pick up Newton, Brookline and the Allston-Brighton part of Boston.

CD 8 takes up the rest of Boston and the near south suburbs including Quincy.

CD 9 keeps Brockton, but otherwise extends across the south suburbs from RI to Mass Bay.

[muon2]

I can't believe you did that.

I've never seen MV and Nantucket split before, but the ferry routes do argue for your division, and it gets the islands double representation.

I would have had to split Barnstable town if I had left Nantucket out of the Cape Cod district.  And then if I had gone from Martha's Vineyard to Wood's Hole and then northward towards the Cape Cod Canal, I would have had to get two districts where there was just one line of towns.  And Plymouth has a fairly substantial population so it might have needed to be split.

So this worked out pretty good population-wise.  I probably could have tried to include the ferry landings in New Bedford, but it ended up that just a small bit of that city, so I took an area adjacent to Fairhaven.

That will be an interesting quandary in MA if splitting the islands keeps a town intact.

[muon2]

I've used the tool to divide Illinois into 18 Congressional districts with the 2008 data. His app uses voting districts which corresponds to the year 2000 precincts. These precincts have changed over the decade, since IL law requires them to adjust to congressional and legislative boundaries after the 2001 remap, and they can split and merge to accommodate growth patterns for each election cycle.

The goal in the map was to create 3 Black and 2 Latino districts in Chicago and suburbs, wiuth detail shown in the second map. The remaining districts were designed to be reasonably compact with minimal numbers of county fragments. No voting data was applied for this version. The maximum population deviation is under 200.

[muon2]

Who does this map pair? It looks like Halvorson and Biggert, would that be accurate?

Where do Quigley and Gutierrez fall in the two earmuff successor districts?

I did some research and saw that Melissa Bean now has a Cook County district and Bill Foster is in the district to the west.

I think the northern earmuff is the Puerto Rican district, and the red district by the stockyards is the Mexican district, so I'd guess Gutierrez has the northern district.   The question then does the Mexican district have enough US adult citizens to be an effective minority majority district.

The standard in Bartlett v. Strickland uses a threshold of 50% of the voting age population. Argument before the SCOTUS raised the question of non-citizens, but the opinion used voting-age population. The app only had population, but I am confident that my percentages are sufficiently above the threshold so the the voting-age standard would also be met.

I started the map by drawing the two Latino districts. New CD 4 (red) is 62% Hispanic and CD 5 (yellow) is 56% Hispanic. New CD 4 gets its additional population from the parts of current CD 3 in Chicago and Berwyn, Stickney and part of Lyons Township in Cook County. New CD 5 takes the northern muff from old CD 4 plus neighboring Chicago areas and suburban Leyden Township from old CD 5 as well as the heavily Hispanic areas in northeast DuPage from current CD 6. I think it is hard to see something significantly different from these two districts drawn without a legal challenge. Gutierrez has talked about retirement in the last couple of cycles, so they could well also both be open in 2012. Quigley's old county district is almost entirely out of this new CD 5.

Next I drew the three Black districts, all of which needed to add substantial population to meet the new district size. New CD 1 (blue) is 53% Black, CD 2 (dark green) is 57% Black, and CD 7 (grey) is 54% Black. CD 7 was drawn first and had to add most of its new population along the lakefront since an expansion into suburban Cook would bring the population dangerously close to 50%. To add enough population new CD 1 picks up White areas of Chicago and Worth Township from current CD 3 and then swaps some with current CD 2 to stay above 50%. CD 2 has the easiest expansion path to the south and easily stays above 50%. As with the Latino districts, and other than some swaps between CDs 1 and 2, I have a hard time picturing any substantial departure from these districts.

With those five districts largely locked in place the other districts and options become clearer. The new CD 3 (purple) is more current CD 13 and both Lipinsky and Biggert live in that district. Given the need for the minority districts, the new CD 3 largely has to move into current 13 or 11 and will likely be a competitive district in 2012. Halvorson actually lives in CD 2 as I've drawn it, but a move elsewhere in the district would not be hard. I've also drawn both Quigley and Schakowsky into the new CD 9 (light blue), though I could see a gerrymander that puts Schkowsky and the city of Evanston in the new CD 10 (pink) and moves CD 5 out to the suburbs by O'Hare.

[muon2]

My take on CO tried to minimize county fragments while bring population deviations down to 100 or less. Presumably, adjustments at the block level could get perfect equality without significant change to the map. While I only split 4 counties while maintaining some compactness, I also note that the result is remarkably similar to the current map. The major difference is that Aurora and Arapahoe County move to CD 7 from 6 and 1, and Lakewood moves into CD 6 from 1.



Edit: I switched to a split of El Paso as suggested in the comments that followed this original post.

[muon2]

My take on CO tried to minimize county fragments while bring population deviations down to 100 or less. Presumably, adjustments at the block level could get perfect equality without significant change to the map. While I only split 4 counties while maintaining some compactness, I also note that the result is remarkably similar to the current map. The major difference is that Aurora and Arapahoe County move to CD 7 from 6 and 1, and Lakewood moves into CD 6 from 1.

This is a good map.  I'd drop the split of Lake County, you can't get from Lake to Pitkin in the winter.  Go ahead and take a few 1000 from southern El Paso instead.

I've driven Independence Pass (12,093 ft) between Piktin and Lake so I know what you are saying about the connection. Actually, my preference was to split Custer, but the voting districts didn't provide fine control to get to my target of a 100 person deviation. With block-level projections, that's probably where I'd go.

[muon2]

This is a good map.  I'd drop the split of Lake County, you can't get from Lake to Pitkin in the winter.  Go ahead and take a few 1000 from southern El Paso instead.

I've driven Independence Pass (12,093 ft) between Piktin and Lake so I know what you are saying about the connection. Actually, my preference was to split Custer, but the voting districts didn't provide fine control to get to my target of a 100 person deviation. With block-level projections, that's probably where I'd go.

My quality measure for a county split would be the relative size of the smaller fragment to the population of the county.  I'd still go for a couple thousand from El Paso rather than splitting either Lake or Custer counties.

The election officials in larger counties are going to be able to handle multiple legislative districts, and probably have an integrated GIS system to draw updated precincts.  A smaller county might have hand drawn maps, and the only reason they have precincts is to match up with the comissioner districts.

In that case, I would keep Park intact and use a bit of NW El Paso for CD 6 (teal) instead.

[muon2]

If you were to reduce the blue district to 51%-52% African-American, how might that tidy up the lines? Not that it isn't an impressively acceptable gerrymander as it is... Is there any way to trade the SE corner to that district so the yellow district doesn't extend that far south?

It's possible, but unfortunately my map didn't save in the app so I don't have it available to play around with.

My guess is that it would be hard to move the SE corner since Houston County is roughly 100 K. You would have to see what that does to your CD 2 percentages.

[muon2]

If you were to reduce the blue district to 51%-52% African-American, how might that tidy up the lines? Not that it isn't an impressively acceptable gerrymander as it is... Is there any way to trade the SE corner to that district so the yellow district doesn't extend that far south?

It's possible, but unfortunately my map didn't save in the app so I don't have it available to play around with.

My guess is that it would be hard to move the SE corner since Houston County is roughly 100 K. You would have to see what that does to your CD 2 percentages.

I took a look at the AL map on the App, and found a way to solve the SE corner issue, keep both black-majority CDs, and minimize county splits to make nicer district lines. All districts are within 200 persons of the ideal size. The Birmingham district is 57% black and the Montgomery-Mobile district is 52% black. CD 1 connects western Mobile County to the rest of the district through Dauphin Island and the ferry to Fort Morgan in Baldwin County.

[muon2]

The other interesting task is to see how close in population the districts can be without splitting any counties. MS has lots of counties and few districts, so it's possible to get quite close. This is my version maintaining one black district. The four deviations are -339, +143, -165, +363. Note that the map converter has obscured the image somewhat, and Walthall and Marion Counties are in CD 3 (purple)

[muon2]

Given the results of the 2008 election I wonder if VRA districts are really even necessary in some parts of the country.  I'd wager that in most places a minority candidate wouldn't have too much trouble getting elected in a 35% minority district.  If the minority votes as a bloc for their candidate it would take less than 25% of the white vote to reach 50%.  I honestly think we've reached a point where VRA districts don't really need to be majority-minority districts anymore.  Perhaps 35% is a little optimistic but I think we could realistically move the "requirement" from 50% to 40% and there would be no decrease in the number of minority Congresspeople.  In fact, I would be slightly surprised if there wasn't an increased number of minorities in Congress if states were permitted to draw their districts that way.  Instead of packing all the minorities into one district they could be spread a little thinner giving them more opportunities for success.

That may be true but the SCOTUS ruled otherwise this spring in Bartlett v Strickland. There's no obligation to protect districts with less than 50% of a minority group within. If there is 50% or more available the need to protect them is subject to the Gingles test. The test is such that most states will assume that the test applies and create districts that will pass the VRA with that test. With the Bartlett decision, that will mean 50% or more in a district.

The maps I created for IL, AL, and MS on this thread have all been consistent with my understanding of the VRA since Bartlett.

[muon2]

Illinois with 18 districts:



Click to embiggen.

IL-01 (dark blue, Bobby Rush - D) - Extends southwest to Will County, but remains majority-black, if only just barely (52%).
IL-02 (dark green, Jesse Jackson Jr. - D) - Same as IL-01, except it takes in more of Will County and is 53% black.
IL-03 (dark purple, Dan Lipinski - D and Judy Biggert - R) - Extends into the DuPage County suburbs, taking in Judy Biggert's home. The population is still centered in Cook County, so Lipinski should be safe, but it's a more appropriate district for his moderate views.
IL-04 (dark red, Luis Gutierrez - D) - Somewhat of a gerrymander still, but much less so than before. I hope Gutierrez doesn't live in the old northern part of the district. 70% Hispanic.
IL-05 (yellow, Michael Quigley - D) - Didn't change this one much. It's only 54% white.
IL-06 (teal, Pete Roskam - R) - Now stretches up from DuPage around to take in some Republican parts of northern Cook and Lake Counties. Probably somewhat more Republican now.
IL-07 (grey, Danny Davis - D) - I had a hell of a time getting three black-majority districts out of Illinois. This one is 52% black. I think we'll be seeing the loss of one of the black-majority districts and gaining another Hispanic-majority district in Illinois either in 2010 or 2020.
IL-08 (light purple, Melissa Bean - D) - Reconfigured the district to drop McHenry and add more of northwestern Cook County. Should be more Democratic.
IL-09 (very light blue, Jan Schakowsky - D) - Pretty unchanged, solidly Dem district.
IL-10 (magenta, Mark Kirk - R) - Pretty much unchanged, so it should still be a Democratic district. If only a Democrat can win it in 2010.
IL-11 (very light green, Will County-based district, Debbie Halvorson - D) - Shrunk this district to just Will and Kendall Counties. Should be easy for Halvorson to hold.
IL-12 (very light purple in the southwest, Jerry Costello - D) - More or less unchanged, added some swing counties in the north of the district and dropped some Republican parts in the southeast.
IL-13 (pink, John Shimkus - R) - Formerly the 19th district, pretty much all Republican territory in the south of the state.
IL-14 (brown, Bill Foster - D) - Replacing the phallic old district, IL-14 now stretches from Foster's home base of Aurora/Batavia, through DuPage and up to Rockford. Should be more Democratic now.
IL-15 (orange, Timothy Johnson - R) - Created a slightly Republican-leaning district that should be competitive in an open seat, but Johnson probably wouldn't break a sweat holding.
IL-16 (light green district in the northwest corner, Don Manzullo - R) - Takes in all the swingy and Republican territory in the northwest, should be safe for Manzullo.
IL-17 (purple district with spidery tendrils, Phil Hare - D) - Still a gerrymander but much less ridiculous; shouldn't change the partisan composition much.
IL-18 (yellow, Aaron Schock - R) - Takes in all the Republican territory in the middle of the state. Safe Republican.

How does it survive a court challenge? The map fails to create a second majority-Hispanic district in Chicago/Cook yet there are demonstrably enough Hispanic adults in close proximity to justify the creation of one under section 2 of the VRA.

I point this out because this is what puts the squeeze on CD-3 and 5. I'm curious to see your Dem gerrymander to protect those D-incumbents yet comply with the VRA. I went for compactness in my map below and didn't look at incumbent protection, but you could take it a different direction.

[muon2]

Quigley's district could be made Hispanic-majority with just a little bit of fiddling along the border with Gutierrez's district.

When I looked at it it took more than a little fiddling. As JL points out it's hard balancing 3 black and 2 hispanic districts. I agree that IL-5 becomes the second majority-hispanic district, but it probably needs that entire northern lobe of IL-4 to do it. That pushes IL-4 into IL-3 which is why I talked about a squeeze there.

[muon2]

I just tried to play around with the borders of my map, and I managed to get Quigley's district up to 44% Hispanic before I gave up, because there's nowhere else I could take territory from except for IL-7, which would have ended up going under 50% black. So while it's possible to make another Hispanic-majority district, it's still difficult, and I doubt they'd enforce VRA in this case.

That's why I added my map - to show that it was possible without extreme gerrymandering. If a map is presented that does not create the five minority districts, there would be a court challenge. I can't find an argument that a federal judge would support allowing only four minority-majority districts.

I solved the IL-7 problem by taking it out to the western edge of Cook where there is a significant black population in Proviso township. But it looks like you use parts of Proviso to keep IL-3 sufficiently Democratic, and use the rest to keep IL-5 from moving to the north. I think the Dems have a real challenge to protect IL-3, since I think the VRA section 2 argument for IL-4 and 5 will be hard to overcome.

[muon2]

In response to the Jim Crow series, here's an anti-Jim Crow map for SC. It uses the 2008 data to create 7 CDs since SC may gain one. All seven districts are within 100 of the ideal population. Both CD-6 (teal) and CD-7 (grey) have just over 50% black population.