Wisconsin Redistricting, 2010

[Verily]

Wisconsin doesn't lose a seat, so it's boring. Paul Ryan is kind of being screwed by changing voting patterns in his area, but he won with two-thirds of the vote in 2008 while Obama carried his district, so the Democrats would have to get their act together to actually be a threat there. Otherwise, the districts are all safe for their incumbents. For the most part, the districts are very similar to the 2000 districts.

I allowed slightly more variation in district population to avoid splitting counties; the least ideal district is WI-08, which is .69% undersized. WI-04 is .63% oversized, while WI-05 is .57% oversized. Everything else is within 0.5% of ideal. As usual, the populations are extrapolated to 2010 using the 2007-2008 growth rate, compounded. Only two counties are split, Milwaukee and Manitowoc; Milwaukee was unavoidable due to size, and Manitowoc was necessary to keep the districts compact.

 (Edited version)

[Padfoot]

This is a great series Verily, keep it up.

[Verily]

Unfortunately, there are relatively few states left with both a substantial number of districts and reasonable counties which allow for county-based redistricting. Racial redistricting also compounds the problem because it means tons of otherwise unnecessary county splits.

I think I'm going to try Georgia next, leaving Minnesota to the people on the thread devoted to it. Georgia can generally do racial districting without too many county splits. After that, Indiana, then maybe Michigan or Ohio. Illinois has too many county splits in the Chicago area to be interesting, although I might do it later.

Current plan:

1. New York (upstate only)
2. Iowa
3. Missouri
4. Wisconsin
5. Georgia
6. Indiana
7. Michigan
8. Ohio
9. Arkansas
10. Oklahoma
11. Louisiana
12. New Mexico
13. Utah
14. Colorado
15. Virginia? (independent cities ugh)
16. West Virginia?
17. Tennessee?
18. Illinois?
19. Oregon?
20. Washington?

[muon2]

What population numbers are you using for these splits? I keep splits for all 50 states based on latest estimates projected ahead to April 1, 2010. For instance I can get all the WI districts within 0.5% while only splitting Milwaukee County. Depending on the municipal population estimates for Milwaukee County released later this summer it could be better - within 0.2%.

[Verily]

What population numbers are you using for these splits? I keep splits for all 50 states based on latest estimates projected ahead to April 1, 2010. For instance I can get all the WI districts within 0.5% while only splitting Milwaukee County. Depending on the municipal population estimates for Milwaukee County released later this summer it could be better - within 0.2%.

Is there a good website for uploading Excel files? I could upload my data sheet. My estimated 2010 population for Wisconsin is 5,686,569 people.

I'm using data from the Census county estimates: http://www.census.gov/popest/counties/CO-EST2008-03.html

Like I said in the initial post, I could have avoid the Manitowoc split. But it would have forced me to split up like areas and create an unreasonably shaped district, which I consider much more important than following county lines. County lines are arbitrary across most of the country, and enslaving yourself to them often results in poorly designed districts. The idea of the Manitowoc split is that Two Rivers and Mishicot go in WI-08 while Manitowoc itself goes in WI-06.

As for the unevenness of the splits, it would be possible to get them slightly closer to 0.5% off. For example, right now WI-08 is 99.31% of ideal while WI-06 is 99.77% of ideal. So WI-08 could be brought to within 0.5% of ideal by changing how I split up Manitowoc County without making WI-06 more than 0.5% beyond the ideal. (This is not true of WI-04 and WI-05 and the Milwaukee County split since WI-04 is 100.64% of ideal and WI-05 is 100.57% of ideal.)

Finally, it's worth pointing out that Census estimates are not 100% accurate, and the degree to which they are off is likely to be greater than the 0.5% range for districts in any case. So, no matter how precise the districts are, some would have to change once the actual 2010 Census numbers are released.

[Hash]

Try box.net. That's a great site.

[muon2]

What population numbers are you using for these splits? I keep splits for all 50 states based on latest estimates projected ahead to April 1, 2010. For instance I can get all the WI districts within 0.5% while only splitting Milwaukee County. Depending on the municipal population estimates for Milwaukee County released later this summer it could be better - within 0.2%.

Is there a good website for uploading Excel files? I could upload my data sheet. My estimated 2010 population for Wisconsin is 5,686,569 people.

I'm using data from the Census county estimates: http://www.census.gov/popest/counties/CO-EST2008-03.html

Like I said in the initial post, I could have avoid the Manitowoc split. But it would have forced me to split up like areas and create an unreasonably shaped district, which I consider much more important than following county lines. County lines are arbitrary across most of the country, and enslaving yourself to them often results in poorly designed districts. The idea of the Manitowoc split is that Two Rivers and Mishicot go in WI-08 while Manitowoc itself goes in WI-06.

As for the unevenness of the splits, it would be possible to get them slightly closer to 0.5% off. For example, right now WI-08 is 99.31% of ideal while WI-06 is 99.77% of ideal. So WI-08 could be brought to within 0.5% of ideal by changing how I split up Manitowoc County without making WI-06 more than 0.5% beyond the ideal. (This is not true of WI-04 and WI-05 and the Milwaukee County split since WI-04 is 100.64% of ideal and WI-05 is 100.57% of ideal.)

Finally, it's worth pointing out that Census estimates are not 100% accurate, and the degree to which they are off is likely to be greater than the 0.5% range for districts in any case. So, no matter how precise the districts are, some would have to change once the actual 2010 Census numbers are released.

Our statewide projections are close. When I add up the county projections I get 5.6874 K which is about 800 persons more than yours so that is certainly within any projection error. Here's the map I get:



Other than the awkward line around Shawano it seems reasonably compact. These are my district populations:

CD 1 (Racine) 710.5 K (-0.05%)
CD 2 (Madison) 711.1 K (+0.02%)
CD 3 (Eau Claire) 711.0 K (+0.01%)
CD 4 (Milwaukee) 709.5 K (-0.20%)
CD 5 (Waukesha) 709.0 K (-0.27%)
CD 6 (Oshkosh) 712.3 K (+0.20%)
CD 7 (Wausau) 711.9 K (+0.14%)
CD 8 (Green Bay) 712.0 K (+0.15%)

[Verily]

Ah, it's mostly because you split Milwaukee County three ways. That makes sense. I also don't like that shape of WI-01, which is mostly suburban and urban but then contains a long rural tail off to the west. (Shawano is obvious.) I'm not sure avoiding the split of Manitowoc is worth the sacrifices that map makes, although I could further optimize populations of the districts using the three-way Milwaukee split.

Edit: I optimized the populations, bringing everything within 0.5% Except that I just noticed a minor math error that means my WI-02 is just a hair below the .5% threshold at 99.47% of ideal. It's fixable by switching Waushara and Green Lake Counties between WI-02 and WI-06, although the new map is a bit less compact.

New version (also on the first post):

[muon2]

Ah, it's mostly because you split Milwaukee County three ways. That makes sense. I also don't like that shape of WI-01, which is mostly suburban and urban but then contains a long rural tail off to the west. (Shawano is obvious.) I'm not sure avoiding the split of Manitowoc is worth the sacrifices that map makes, although I could further optimize populations of the districts using the three-way Milwaukee split.

Except for Janesville and Beloit pretty much everything west of Racine and Kenosha is rural with small towns. I don't see that attaching Green/Iowa/Lafayette to Rock is any different than attaching them to Dane.

I count county fragments as being a part of a county that is not a whole district. My map has two such fragments versus three in yours.

[Verily]

My count for your Green Bay district ends up at +2.04%, not +0.15% as you have. I'm not sure what the discrepancy is. Never mind, figured it out. Although now I have it at +0.01% (actually +0.0059%).

[muon2]

My count for your Green Bay district ends up at +2.04%, not +0.15% as you have. I'm not sure what the discrepancy is. Never mind, figured it out. Although now I have it at +0.01% (actually +0.0059%).

The difference seems consistent with our slightly different projections as seen in the state total.

[Verily]

Anyway, I did another revision to fix a mistake I'd made (used Polk's population for Pierce), which fortunately was a difference of only some 3,000 people and thus easily rectified. Anyway, my new map looks a lot more like yours, although I don't have the awkwardness around Shawano County.

[muon2]

Anyway, I did another revision to fix a mistake I'd made (used Polk's population for Pierce), which fortunately was a difference of only some 3,000 people and thus easily rectified. Anyway, my new map looks a lot more like yours, although I don't have the awkwardness around Shawano County.

I couldn't use that map since it puts the Green Bay district (WI-08) at +0.588% which is outside my tolerance. I assume that this is due to a difference in our projections. For my projections I get 715.1 K with a target of 710.9 K population.

[Verily]

We must have something really different in there, because I have -0.13% for that district. 709,863 people against an ideal of 710,821 people.

Manitowoc: 80,388
Brown: 248,274
Door: 27,811
Kewaunee: 20,239
Outagamie: 177,796
Waupaca: 51,702
Shawano: 40,972
Menominee: 4,516
Langlade: 19,941
Oconto: 37,734
Total: 709,863

(Marinette: 41,972)

Do you use single-year numbers for your projection (that is, 2007-2008), or the average annual growth 2000-2008? If it's the latter, perhaps Shawano County was growing quickly at one point but has now stagnated (it lost a single person 2007-2008, which for my purposes is counted as a growth rate of 0%), which would throw off the long-term projection.

Otherwise, all sources seem to say that Shawano County is smaller than Marinette County, so I don't see how switching them could possibly give a larger population (although Marinette is declining slightly while Shawano is stagnant).

[muon2]

We must have something really different in there, because I have -0.13% for that district. 709,863 people against an ideal of 710,821 people.

Manitowoc: 80,388
Brown: 248,274
Door: 27,811
Kewaunee: 20,239
Outagamie: 177,796
Waupaca: 51,702
Shawano: 40,972
Menominee: 4,516
Langlade: 19,941
Oconto: 37,734
Total: 709,863

(Marinette: 41,972)

Do you use single-year numbers for your projection (that is, 2007-2008), or the average annual growth 2000-2008? If it's the latter, perhaps Shawano County was growing quickly at one point but has now stagnated (it lost a single person 2007-2008, which for my purposes is counted as a growth rate of 0%), which would throw off the long-term projection.

Otherwise, all sources seem to say that Shawano County is smaller than Marinette County, so I don't see how switching them could possibly give a larger population (although Marinette is declining slightly while Shawano is stagnant).

I use the following method for my projections. The 2008 estimate and the 2000 census figures are used to calculate an annual rate (compounded) based on 8.25 years. The extra quarter year is the difference between the April 1 census and the July 1 estimate dates. That rate is applied to the 2000 census base over 10 years to get the 2010 projection.

I do have Marinette (42,059) projected to be about 1000 larger than Shawano (41,038). I must have shifted counties incorrectly since I agree that it makes the Green Bay district closer to ideal, however it shifts my Wausau district even higher, and my current map is the optimum between those two districts. If I make your other shifts between Eau Claire and Wausau then I get Eau Claire up to +0.46%, within tolerance, but not as well balanced in pop.