Wisconsin Legislative Redistricting

[muon2]

Here's the rest of northern WI. Of the cities not previously mentioned only Eau Claire is chopped and the chop follows the Chippewa and Eau Claire rivers, which makes vote calculations uniform from 2008 to 2012. All of Wausaw and Schofield is in AD 83, which may not be clear from the map detail.



AD 82 (Shawano): deviation +163; PVI R+4 (-1.0, -6.8 )
AD 83 (Wausaw): deviation +2324; PVI D+1 (2.9, -1.9)
AD 84 (Weston): deviation +1193; PVI R+5 (-1.3, -7.9)

AD 85 (Merinette): deviation -1968; PVI R+2 (-0.5, -4.1)
AD 86 (Rhinelander): deviation -1469; PVI R+2 (0.0, -4.7)
AD 87 (Merrill): deviation +2015; PVI D+0 (1.9, -1.8 )

AD 88 (Ashland): deviation +1073; PVI D+4 (4.5, 2.6)
AD 89 (Superior): deviation +2626; PVI D+9 (9.1, 8.9)
AD 90 (Osceola): deviation +2218; PVI R+5 (-4.1, -6.2)

AD 91 (Chippewa Falls): deviation +2664; PVI R+1 (0.2, -2.7)
AD 92 (Eau Claire - north): deviation -780; PVI D+7 (8.1, 5.9)
AD 93 (Eau Claire - south): deviation +1690; PVI D+4 (5.8, 2.9)

AD 94 (Medford): deviation -2065; PVI R+5 (-1.4, -8.7)
AD 95 (Black River Falls): deviation -1255; PVI D+6 (8.1, 4.3)
AD 96 (River Falls): deviation -2293; PVI R+0 (1.4, -1.8 )

AD 97 (Rice Lake): deviation +1458; PVI R+1 (-0.0, -1.6)
AD 98 (Menomonee): deviation +293; PVI R+2 (0.5, -4.8 )
AD 99 (Hudson): deviation -11; PVI R+7 (-5.8, -8.1)

In this group there are 3 D, 2 d, 5 e, 7 r, 1 R.

For the whole plan there are 35 solid D, 12 lean d, 13 even, 16 lean r, and 23 solid R.

[muon2]

Here's some preliminary analysis of my AD plan. The distribution 35D-12d-13e-16r-23R gives a 47 to 39 advantage to the Dems in the PVI of the districts which is an 8 seat margin. The PVI of the state was D+2.4 (3.4, 1.5). Studies predict a 4% swing in seat margin for every 1% change in state PVI, so the expected margin for the Dems would be 10 seats. Thus the plan has a SKEW of 2 to the Pubs for the 99 seats. Obama won 74 of these districts in 2008 and 57 in 2012, so it seems very close to fair in a partisan sense.

The plan has 41/99 districts that are lean or even which is better than the 30% that is expected. The POLARIZATION is 144. 17 seats shifted from 2008 to 2012 even though Obama's statewide 2-party vote only dropped from 57.1% to 53.5%, which is twice the number expected. That points to a reasonably competitive plan that would be responsive to changes in voter preference.

The above analysis is based on both 2008 and 2012 data. One interesting result was the PVI shift between the two years. The two point swing R isn't so unusual, but what is more telling is that Milwaukee and Madison swung D, while the rest of that state moved further R than the statewide number suggests. That increased polarization shows up as I look at the score based on each year separately. It goes from 137 in 2008 to 156 in 2012. With that concentration, the turnout in Milwaukee and Madison becomes more critical for the Dems. It seems to presage the results of 2016 as the outstate areas continued to polarize towards the Pubs.

[Gass3268]

Do you have a map showing the Janesville split?

[muon2]

Do you have a map showing the Janesville split?

Janesville city is completely whole (AD 10) except for ward 5 on DRA which is the far northern end of the city. That corresponds essentially to wards 19, 20, 21, 22 and 26 on the current ward map.

[Gass3268]

Do you have a map showing the Janesville split?

Janesville city is completely whole (AD 10) except for ward 5 on DRA which is the far northern end of the city. That corresponds essentially to wards 19, 20, 21, 22 and 26 on the current ward map.

Thanks, but it was jimrtex's map where I couldn't tell where the split was in Janesville. Once I'm done with his map, I'll go through your's to come up with the 2016 numbers.

[muon2]

Do you have a map showing the Janesville split?

Janesville city is completely whole (AD 10) except for ward 5 on DRA which is the far northern end of the city. That corresponds essentially to wards 19, 20, 21, 22 and 26 on the current ward map.

Thanks, but it was jimrtex's map where I couldn't tell where the split was in Janesville. Once I'm done with his map, I'll go through your's to come up with the 2016 numbers.

Great. I would do it myself, but I can't do the chopped cities since Atlas doesn't have ward results posted yet. I am especially interested in the vote totals so that I can calculate the wasted votes and efficiency gap as I'll discuss in an upcoming post.

[jimrtex]

I have decided to eliminate the single-district counties with a deviation of between 5% and 10% (Sauk, Waupaca, and Chippewa). Rather than redo the entire map, I made as few changes as possible - as if I was responding to a court decision.

Sauk (1.079) is added to the region comprised of Dane, Green, Iowa, and Lafayette. The average error for that group improves from -1.6% to -0.7%.

(7) Dane, Sauk, Green, Iowa, and Lafayette 10.922 (11), -0.7% 8 districts in Dane, 1 district in Sauk, 1 in Green and Dane(part) (65% in Green), and 1 in Iowa, Lafayette, Sauk(part), and Dane (part)(8/10 in Iowa, Lafayette, and Sauk). Three surplus fragments, with two surplus fragments in Dane.

Waupaca (0.912) is added to Outagamie, recognizing a small surplus in Outagamie. The average error improves from +2.5% to -0.3%.

(12) Outagamie and Waupaca, 3.988 (4) -0.3%. Three districts in Outagamie, with the fourth district comprised of Waupaca (90% plus of district) and a small surplus fragment of Outagamie. One surplus fragment.

Chippewa (1.087) is added to a region comprised of Lincoln, Taylor, Sawyer, Washburn, Rusk, and Price 1.929(2) -3.5%. The region is then split, with a northern region consisting of Sawyer, Washburn, Rusk, and Price; and southern region comprised of Chippewa, Lincoln, and Taylor. The northern district will be slightly overpopulated, but a split of a small county is eliminated.

(17) Chippewa, Lincoln, and Taylor, 1.947 (2) -2.6%. One district in Chippewa, one district in Lincoln and Taylor, with a small portion of Chippewa (about 1/9 of the district). One surplus fragment.

(27) Sawyer, Washburn, Rusk, and Price 1.069(2) +6.9%. One district, no county splits Slightly excessive deviation.



(1) Milwaukee and Ozaukee 18.002(18), +0.0%. 16 districts in Milwaukee, 1 in Ozaukee, 1 straddling the county line. Two surplus fragments.

(2) Sheboygan 2.011(2), +0.5%. 2 districts in county. No surplus fragments.

(3) Waukesha and Washington 9.083(9), +0.9%, 6 districts in Waukesha, 2 in Washington, 1 straddling the line, about 4/5 in Waukesha. Two surplus fragments.

(4) Dodge and Jefferson 3.002(3), +0.1%, 1 district in Dodge, 1 in Jefferson, one straddling the line. Two surplus fragments.

(5) Kenosha 2.897(3), -3.4%, 3 districts in county. No surplus fragments.

(6) Racine, Rock, and Walworth 7.972(8), -0.3%. 3 districts in Racine, 2 in Rock, and 1 in Walworth, one straddling Racine-Walworth line (3/5 in Walworth), and one straddling Rock-Walworth line (4/5 in Rock). Four surplus fragments, including two in Walworth.

(7) Dane, Sauk, Green, Iowa, and Lafayette 10.922 (11), -0.7% 8 districts in Dane, 1 district in Sauk, 1 in Green and Dane(part) (65% in Green), and 1 in Iowa, Lafayette, Sauk(part), and Dane (part)(8/10 in Iowa, Lafayette, and Sauk). Three surplus fragments, with two surplus fragments in Dane.

(8) Grant, Vernon, Richland, and Crawford    2.013(2), +0.7%. One district in Grant and Crawford (part), and one in Vernon, Richland, and Crawford (part). Conceivably the split could be in Richland. One divided small county.

(9) La Crosse 1.996(2), 0.2%. 2 districts in county. No surplus fragments.

(10) Fond du Lac and Green Lake 2.101(2), +5.04%. One district in Fond du Lac, and one straddling the border. One surplus fragment. The excess deviation could be avoided by pairing Fond du Lac with Washington.

(11) Manitowoc, Calumet, Door, and Kewaunee 3.112(3) +3.7%. One district in Manitowoc, one in Calumet and Manitowoc (part), with about 4/5 in Calumet, one in Door, Kewaunee, and Manitowoc (part), with about 4/5 in Door and Kewaunee. Two surplus fragments in Manitowoc.

(12) Outagamie and Waupaca, 3.988 (4) -0.3%. Three districts in Outagamie, with the fourth district comprised of Waupaca (90% plus of district) and a small surplus fragment of Outagamie. One surplus fragment.

(13) Winnebago 2.907 (3) -3.1%. Three districts in county. No surplus fragments.

(14) Columbia 0.989(1) -1.1%. One district in county.

(15) Portage, Waushara, and Marquette 1.913(2) -4.3%. One district in Portage. One in Waushara, Marguette, and Portage(part), with about 7/10 in Waushara and Marquette. One surplus fragment.

(16) Marathon and Shawano 3.064(3) +2.1%. Two districts in Marathon, and one straddling the border with about 7/10 in Shawano. One surplus fragment.

(17) Chippewa, Lincoln, and Taylor, 1.947 (2) -2.6%. One district in Chippewa, one district in Lincoln and Taylor, with a small portion of Chippewa (about 1/9 of the district). One surplus fragment.

(18) Brown and Oconto 4.973(5) -0.5%. Four districts in Brown, and one straddling the border with about 2/3 in Oconto. One surplus fragment.

(19) Marinette, Forest, and Florence   0.966(1) -3.4.%. One district, no county splits.

(20) Oneida, Langlade, and Menominee 1.048(1) +4.8%. One district, no county splits. Conceivably, Menominee could be placed with Shawano or Oconto. This would reduce the deviation range some, but place Menominee in a district that extends into Brown or Marathon counties.

(21) Douglas and Burnett 1.038(1) +3.8%. One district, no county splits.

(22) Wood and Clark 1.905(2) -4.7%. One district in Wood, one straddling the border. about 3/5 in Clark. One surplus fragment.

(23) Vilas, Adams, Bayfield, and Iron 1.019(1) +1.9%. One district, no county splits.

(24) Monroe, Juneau, Adams, and Jackson 1.961(2) -1.9%. Two districts: Juneau and Adams plus part of Monroe (1/5); and Monroe (4/5) and Jackson. One divided small county. D'ja know that Wisconsin does not have a Van Buren County?

(25) St. Croix, Polk, and Pierce 2.952(3) -1.6%. One district in St. Croix, one straddling Polk and St Croix line (4/5 in Polk), and one straddling Pierce and St. Croix line (7/10) in Pierce. Two surplus fragments.

(26) Eau Claire, Barron, Dunn, Trempealeau, Buffalo, and Pepin 4.149(4) +3.7%. One district in Eau Claire; one in Barron and part of Dunn (30% of Dunn); remainder of Dunn and part of Eau Claire; Trempealeau, Buffalo, Pepin and part of Eau Claire. Alternatively, Pepin and Buffalo could be placed with Dunn, switching the distribution of the Eau Claire surplus. My initial inclination is to keep the Mississippi River counties together. Two surplus fragments and one divided small county.

(27) Sawyer, Washburn, Rusk, and Price 1.069(2) +6.9%. One district, no county splits Slightly excessive deviation.

Summary of excessive divisions.

26 Surplus fragments, seven more than necessary. Milwaukee and Waukesha could have 16 and 7 districts respectively, but their fractions of 0.498 and 0.783 are far from being negligible or almost a whole. If we exclude them, there are are 5 surplus fragments. No large county has fewer districts than the whole number it is entitle to, so it is just the case of the surplus being divided between two districts for Walworth, Dane, Manitowoc, St. Croix, and Eau Claire.

Three small counties are split.

[jimrtex]

Do you have a map showing the Janesville split?

Janesville city is completely whole (AD 10) except for ward 5 on DRA which is the far northern end of the city. That corresponds essentially to wards 19, 20, 21, 22 and 26 on the current ward map.

Thanks, but it was jimrtex's map where I couldn't tell where the split was in Janesville. Once I'm done with his map, I'll go through your's to come up with the 2016 numbers.

Working on it.

My map  is based on political  subdivisions: towns, cities, and villages. When I split a city, I use VTD's, but the VTD's don't match the cities. So I'm having to find VTD's that are split between subdivisions. Usually, these are town VTD's that have a little bit of a city that has been annexed, or maybe the VTD's haven't been updated. When this happens I try to figure out the city ward that the population belongs to and add it to the VTD for that ward for the city.

In a few instances there is a small part of a town in a city VTD. In those cases, I simply subtract out the town population.

So the total population of the VTD's for a city (that is divided) will match the population for the city. The boundaries of the VTD's won't precisely match the city limits.

I'm not bothering trying to update all the VTD populations, since for political subdivisions that are not split, I never use the VTD (or ward) except as part of a total for the subdivision.

In real life, a state government should require political subdivisions to delineate neighborhoods. If necessary, they should feed new block boundaries to the Census Bureau.

In December of the '0' year, the Census Bureau releases the census geography (e.g block boundaries, but not block populations). The state would then be able to define neighborhoods in terms of census blocks. Once the PL 94-171 data arrives, the state can go ahead and redistrict. If they have to divide a city, they can divide based on neighborhoods.

An ideal assembly district size is roughly 57,000. Neighborhoods of 5,000 or less can generally be assembled into districts within 5% deviation. Perhaps cities with a population below 10,000 or even 20,000 can be exempted. There would generally be other ways to group communities. If it were absolutely necessary to divide such a small community, it could be done on a special case basis, where the city participates in the division.

[jimrtex]

Do you have a map showing the Janesville split?

Janesville city is completely whole (AD 10) except for ward 5 on DRA which is the far northern end of the city. That corresponds essentially to wards 19, 20, 21, 22 and 26 on the current ward map.

Thanks, but it was jimrtex's map where I couldn't tell where the split was in Janesville. Once I'm done with his map, I'll go through your's to come up with the 2016 numbers.

Finished. I also split off Ward 5, in calculating the population it includes areas that were in VTDs for Harmony town.

[muon2]

I have decided to eliminate the single-district counties with a deviation of between 5% and 10% (Sauk, Waupaca, and Chippewa). Rather than redo the entire map, I made as few changes as possible - as if I was responding to a court decision.

Sauk (1.079) is added to the region comprised of Dane, Green, Iowa, and Lafayette. The average error for that group improves from -1.6% to -0.7%.

(7) Dane, Sauk, Green, Iowa, and Lafayette 10.922 (11), -0.7% 8 districts in Dane, 1 district in Sauk, 1 in Green and Dane(part) (65% in Green), and 1 in Iowa, Lafayette, Sauk(part), and Dane (part)(8/10 in Iowa, Lafayette, and Sauk). Three surplus fragments, with two surplus fragments in Dane.

Waupaca (0.912) is added to Outagamie, recognizing a small surplus in Outagamie. The average error improves from +2.5% to -0.3%.

(12) Outagamie and Waupaca, 3.988 (4) -0.3%. Three districts in Outagamie, with the fourth district comprised of Waupaca (90% plus of district) and a small surplus fragment of Outagamie. One surplus fragment.

Chippewa (1.087) is added to a region comprised of Lincoln, Taylor, Sawyer, Washburn, Rusk, and Price 1.929(2) -3.5%. The region is then split, with a northern region consisting of Sawyer, Washburn, Rusk, and Price; and southern region comprised of Chippewa, Lincoln, and Taylor. The northern district will be slightly overpopulated, but a split of a small county is eliminated.

(17) Chippewa, Lincoln, and Taylor, 1.947 (2) -2.6%. One district in Chippewa, one district in Lincoln and Taylor, with a small portion of Chippewa (about 1/9 of the district). One surplus fragment.

(27) Sawyer, Washburn, Rusk, and Price 1.069(2) +6.9%. One district, no county splits Slightly excessive deviation.

We both have 27 whole county regions, so in principle we should match in chop count. Will you be able to maintain a 10% range given the deviation of (27)? I had to work at mine with only the 5.2% deviation of West Allis. It's grouped with West Milwaukee at -4.7% to get the range to 9.884%. The tricky part was watching some of the lower pop regions to make sure none went beyond the -4.7% deviation. 6.9% forces a floor of -3.1%.

[muon2]

Neutral districts can also say a lot about trends. Since they aren't gerrymandered to pack or crack parties they can show how the natural political leanings are shifting. For example, consider the 6 whole county ADs in northern WI. PVIs reflect a districts political leanings compared to the national presidential average, and I'll look at the last 3 cycles (2008, 2012, 2016) and negative numbers are Pub. Instead of averaging two cycles, I'll keep each cycle separate.



AD 85 (Merinette): PVI (-0.5, -4.1, -18.4)
AD 86 (Rhinelander): PVI (0.0, -4.7, -13.2)
AD 87 (Merrill): PVI (1.9, -1.8, -7.4)

AD 88 (Ashland): PVI (4.5, 2.6, -4.8 )
AD 89 (Superior): PVI (9.1, 8.9, -1.9)
AD 90 (Osceola): PVI (-4.1, -6.2, -15.7)

Notice that the Pub swing was already underway from 2008 to 2012, with shifts of up to almost -5% in AD 86. But the shift for 2016 is really astounding with PVI jumps of -14% in AD 85, and almost -11% in the tradition Dem bastion of Superior. It defies common wisdom that AD 89 would support a Pub candidate, yet Trump won by 458 votes out of over 30K cast. This is one of the clearest indications I've seen of the effect of the Dem loss of the WWC and why Trump won WI.

[jimrtex]

I have decided to eliminate the single-district counties with a deviation of between 5% and 10% (Sauk, Waupaca, and Chippewa). Rather than redo the entire map, I made as few changes as possible - as if I was responding to a court decision.

Sauk (1.079) is added to the region comprised of Dane, Green, Iowa, and Lafayette. The average error for that group improves from -1.6% to -0.7%.

(7) Dane, Sauk, Green, Iowa, and Lafayette 10.922 (11), -0.7% 8 districts in Dane, 1 district in Sauk, 1 in Green and Dane(part) (65% in Green), and 1 in Iowa, Lafayette, Sauk(part), and Dane (part)(8/10 in Iowa, Lafayette, and Sauk). Three surplus fragments, with two surplus fragments in Dane.

Waupaca (0.912) is added to Outagamie, recognizing a small surplus in Outagamie. The average error improves from +2.5% to -0.3%.

(12) Outagamie and Waupaca, 3.988 (4) -0.3%. Three districts in Outagamie, with the fourth district comprised of Waupaca (90% plus of district) and a small surplus fragment of Outagamie. One surplus fragment.

Chippewa (1.087) is added to a region comprised of Lincoln, Taylor, Sawyer, Washburn, Rusk, and Price 1.929(2) -3.5%. The region is then split, with a northern region consisting of Sawyer, Washburn, Rusk, and Price; and southern region comprised of Chippewa, Lincoln, and Taylor. The northern district will be slightly overpopulated, but a split of a small county is eliminated.

(17) Chippewa, Lincoln, and Taylor, 1.947 (2) -2.6%. One district in Chippewa, one district in Lincoln and Taylor, with a small portion of Chippewa (about 1/9 of the district). One surplus fragment.

(27) Sawyer, Washburn, Rusk, and Price 1.069(2) +6.9%. One district, no county splits Slightly excessive deviation.

We both have 27 whole county regions, so in principle we should match in chop count. Will you be able to maintain a 10% range given the deviation of (27)? I had to work at mine with only the 5.2% deviation of West Allis. It's grouped with West Milwaukee at -4.7% to get the range to 9.884%. The tricky part was watching some of the lower pop regions to make sure none went beyond the -4.7% deviation. 6.9% forces a floor of -3.1%.

The overall range is irrelevant, and pandering to innumerate judges.

I will defend my plan based on a meaningful statistic such as standard deviation, and justify the few outliers as avoiding unnecessary division of political subdivisions.

[Gass3268]

Neutral districts can also say a lot about trends. Since they aren't gerrymandered to pack or crack parties they can show how the natural political leanings are shifting. For example, consider the 6 whole county ADs in northern WI. PVIs reflect a districts political leanings compared to the national presidential average, and I'll look at the last 3 cycles (2008, 2012, 2016) and negative numbers are Pub. Instead of averaging two cycles, I'll keep each cycle separate.



AD 85 (Merinette): PVI (-0.5, -4.1, -18.4)
AD 86 (Rhinelander): PVI (0.0, -4.7, -13.2)
AD 87 (Merrill): PVI (1.9, -1.8, -7.4)

AD 88 (Ashland): PVI (4.5, 2.6, -4.8 )
AD 89 (Superior): PVI (9.1, 8.9, -1.9)
AD 90 (Osceola): PVI (-4.1, -6.2, -15.7)

Notice that the Pub swing was already underway from 2008 to 2012, with shifts of up to almost -5% in AD 86. But the shift for 2016 is really astounding with PVI jumps of -14% in AD 85, and almost -11% in the tradition Dem bastion of Superior. It defies common wisdom that AD 89 would support a Pub candidate, yet Trump won by 458 votes out of over 30K cast. This is one of the clearest indications I've seen of the effect of the Dem loss of the WWC and why Trump won WI.

This is true, but I think it's easy to forget just how big and in many ways unprecedented Obama's win was in Wisconsin in 2008. He won counties that Democrats almost never win. 2004 to 2016 would be a better comparison.

[muon2]

I have decided to eliminate the single-district counties with a deviation of between 5% and 10% (Sauk, Waupaca, and Chippewa). Rather than redo the entire map, I made as few changes as possible - as if I was responding to a court decision.

Sauk (1.079) is added to the region comprised of Dane, Green, Iowa, and Lafayette. The average error for that group improves from -1.6% to -0.7%.

(7) Dane, Sauk, Green, Iowa, and Lafayette 10.922 (11), -0.7% 8 districts in Dane, 1 district in Sauk, 1 in Green and Dane(part) (65% in Green), and 1 in Iowa, Lafayette, Sauk(part), and Dane (part)(8/10 in Iowa, Lafayette, and Sauk). Three surplus fragments, with two surplus fragments in Dane.

Waupaca (0.912) is added to Outagamie, recognizing a small surplus in Outagamie. The average error improves from +2.5% to -0.3%.

(12) Outagamie and Waupaca, 3.988 (4) -0.3%. Three districts in Outagamie, with the fourth district comprised of Waupaca (90% plus of district) and a small surplus fragment of Outagamie. One surplus fragment.

Chippewa (1.087) is added to a region comprised of Lincoln, Taylor, Sawyer, Washburn, Rusk, and Price 1.929(2) -3.5%. The region is then split, with a northern region consisting of Sawyer, Washburn, Rusk, and Price; and southern region comprised of Chippewa, Lincoln, and Taylor. The northern district will be slightly overpopulated, but a split of a small county is eliminated.

(17) Chippewa, Lincoln, and Taylor, 1.947 (2) -2.6%. One district in Chippewa, one district in Lincoln and Taylor, with a small portion of Chippewa (about 1/9 of the district). One surplus fragment.

(27) Sawyer, Washburn, Rusk, and Price 1.069(2) +6.9%. One district, no county splits Slightly excessive deviation.

We both have 27 whole county regions, so in principle we should match in chop count. Will you be able to maintain a 10% range given the deviation of (27)? I had to work at mine with only the 5.2% deviation of West Allis. It's grouped with West Milwaukee at -4.7% to get the range to 9.884%. The tricky part was watching some of the lower pop regions to make sure none went beyond the -4.7% deviation. 6.9% forces a floor of -3.1%.

The overall range is irrelevant, and pandering to innumerate judges.

I will defend my plan based on a meaningful statistic such as standard deviation, and justify the few outliers as avoiding unnecessary division of political subdivisions.

Then why change from your original model using single counties with deviations up to 10%? How does that differ in complying with range requirements? In any case, this exercise arose in response to a court decision that WI had to redraw its districts. A plan that doesn't meet the basic court requirements isn't going to be as useful in showing the court what types of alternatives exist to replace the current plan.

[muon2]

Neutral districts can also say a lot about trends. Since they aren't gerrymandered to pack or crack parties they can show how the natural political leanings are shifting. For example, consider the 6 whole county ADs in northern WI. PVIs reflect a districts political leanings compared to the national presidential average, and I'll look at the last 3 cycles (2008, 2012, 2016) and negative numbers are Pub. Instead of averaging two cycles, I'll keep each cycle separate.



AD 85 (Merinette): PVI (-0.5, -4.1, -18.4)
AD 86 (Rhinelander): PVI (0.0, -4.7, -13.2)
AD 87 (Merrill): PVI (1.9, -1.8, -7.4)

AD 88 (Ashland): PVI (4.5, 2.6, -4.8 )
AD 89 (Superior): PVI (9.1, 8.9, -1.9)
AD 90 (Osceola): PVI (-4.1, -6.2, -15.7)

Notice that the Pub swing was already underway from 2008 to 2012, with shifts of up to almost -5% in AD 86. But the shift for 2016 is really astounding with PVI jumps of -14% in AD 85, and almost -11% in the tradition Dem bastion of Superior. It defies common wisdom that AD 89 would support a Pub candidate, yet Trump won by 458 votes out of over 30K cast. This is one of the clearest indications I've seen of the effect of the Dem loss of the WWC and why Trump won WI.

This is true, but I think it's easy to forget just how big and in many ways unprecedented Obama's win was in Wisconsin in 2008. He won counties that Democrats almost never win. 2004 to 2016 would be a better comparison.

Actually the changes are just as striking if one goes back to 2004. The Rhinelander AD I drew was R+2.4 in 2004, and only shifted to even in 2008, but then goes to R+4.7 and R+13.2 in 2012 and 2016. The Superior AD was actually D+12.8 in 2004 and lost some ground in 2008, before careening to R+1.9 in 2016. Both districts (one heavy D and one lean R) shifted Pub by double digits over that period. The beauty of the PVI is that it accounts for the effects of a national wave like Obama in 2008 so one can see the underlying local trends.

[jimrtex]

We both have 27 whole county regions, so in principle we should match in chop count. Will you be able to maintain a 10% range given the deviation of (27)? I had to work at mine with only the 5.2% deviation of West Allis. It's grouped with West Milwaukee at -4.7% to get the range to 9.884%. The tricky part was watching some of the lower pop regions to make sure none went beyond the -4.7% deviation. 6.9% forces a floor of -3.1%.

The overall range is irrelevant, and pandering to innumerate judges.

I will defend my plan based on a meaningful statistic such as standard deviation, and justify the few outliers as avoiding unnecessary division of political subdivisions.

Then why change from your original model using single counties with deviations up to 10%? How does that differ in complying with range requirements? In any case, this exercise arose in response to a court decision that WI had to redraw its districts. A plan that doesn't meet the basic court requirements isn't going to be as useful in showing the court what types of alternatives exist to replace the current plan.

I don't view my plan as a remedial plan. I view it as an exemplar for a case that is still being litigated.

The target is to have districts of 57444.30303 persons.  It is not to have districts of between 54,573 and 60,316 persons. But at the same time we want to avoid unnecessary division of counties, towns, cities, and villages.

We can better demonstrate a good-faith effort attempt to meet the twin constraints using statistical measurements such as standard deviation.

Take a look at what happens when you use the range as a "target". For example, Sauk County is entitled to 1.079 representatives. If you use the range, you can skim between 1666 and 7410 persons and be between 1.05 and 0.95 persons in that district, but if you are near the outer limits you aren't trying to hit the target, and you are engaging in the corruption that occurred in Arizona in 2010, or in Alabama in 2000.

Instead the target for the district entirely in Sauk should be 0.993, the target for the region. That is, once we have accepted a need for county division, we should seek equality within the region, again avoid splitting of towns, cities, and villages as much as possible.

The 10% rule for a single county was something I added when I calculated the original population entitlements, and noticed that there were four counties within 10%, but only one within 5%. It wasn't anything that was really necessary, and I thought it might detract from understanding my overall scheme.

While I said that I was responding to a court decision, that was a lazy fiction. What I really did was make a quick adjustment that didn't require me to redo my overall regional plan. In all three cases, I simply added the single county to another region, which in the case of Chippewa permitted the region to be split.

I announced it when I did because I did not want to later have to redo the Dane region. I did not even look at the Dane region before adding Sauk County. That is, I did not add Sauk in response to difficulties in drawing the map. I had always planned to come into Dane from both Green and Iowa counties.

I think it might be practical to have local redistricting processes. Vermont permits local towns to adjust district boundaries. Global solutions are much more complex, and I don't think yield better results.

There is no 10% overall range requirement. Deviations outside 5% need to be justified, and there are ample court decisions recognizing that my deviations are justifiable.

[muon2]

I find relatively few court decisions that would justify a range in excess of 10%. What you would have to show is that given your goals, there is no way to achieve them without exceeding 10%. That requires a clear statement of those goals, in a way that is not unduly arbitrary, so that one can test them to see if they can be achieved without the deviation.

At this point your goals seem to be:
1) minimize county chops by establishing a maximum number of regions;
2) minimize chops of munis within a region;
2) minimize the standard deviation of the population of the districts.

Feel free to correct this or add as necessary.

[jimrtex]

Dane, Sauk, Green, Iowa, and Lafayette are entitled to 10.922 districts, or 11 whole districts with an  average population of 0.993. Eight districts will be wholly within Dane, one district will be wholly within Sauk, one district will include all of Green and a portion of Dane (roughly 65% in Green), and the other district will combine Iowa, Lafayette, the remnant of Sauk and small portion of Dane (roughly 80% in Iowa, Lafayette, and Sauk).

Initially, I determined the area of Sauk to be attached to Iowa and Lafayette. Next I determined the areas of Dane to be attached to Iowa and Green county. Finally, I divided Dane into 8 districts. Madison city is entitled to 4.060 districts, and including the enclave of Shorewood Hills village, 4.087. This can be divided into four districts, with four districts for the surrounding suburban areas.

While drawing the suburban districts, I adjusted the boundary  for the Green County based district. Originally it had kept to the southern tier of towns. But this made the suburban districts less compact and more unequal. Swapping Verona for Oregon, makes the suburban districts more compact and equal in the population.







My original plan was to include a strip in Dane along the northern edge of Green County to complete the district, but that squeezed the district based in Fitchburg. This configuration swapped Verona for Oregon, which pushes the district closer to Madison, but 2/3 of the district is in Green.

45. Dane: Belleville village, Montrose town, Perry town, Primrose town, Springdale town, Verona city, and Verona town. 0.328; Green (all) 0.641. 0.970.

AD-45 will be paired with two Rock County districts in a senate district.

I identified towns in the southern part of Sauk County that when removed made the Sauk county district at the regional average. I thin added Mount Horeb from Dane County to get to the overall target.

46. Dane: Blue Mounds town, Blue Mounds village, Mount Horeb village, and Vermont town. 0.168; Iowa (all) 0.412; Lafayette (all) 0.293; Sauk: Bear Creek town, Franklin town, Plain village, Spring Green town, Spring Green village, and Troy town. 0.107. 0.980.

47. Sauk (all except that in AD-46) 0.972.

I had originally planned to place the Sauk County district into a senate district with assembly districts to the west. This would avoid a senate chop of Sauk County. But now that Sauk is divided, I will place AD-46 withthe area to the west (Grant, Vernon, Richland and Crawford) in a fairly rural district.

The eight Dane districts will be augmented by the Columbia district to form three senate districts. AD-47 will be joined to districts going northward.

[muon2]

So you are allowing connections across lakes? Both Monona and Maple Bluff are surrounded by Madison except for frontage on a lake.

[muon2]

Neutral districts can also say a lot about trends. Since they aren't gerrymandered to pack or crack parties they can show how the natural political leanings are shifting. For example, consider the 6 whole county ADs in northern WI. PVIs reflect a districts political leanings compared to the national presidential average, and I'll look at the last 3 cycles (2008, 2012, 2016) and negative numbers are Pub. Instead of averaging two cycles, I'll keep each cycle separate.



AD 85 (Merinette): PVI (-0.5, -4.1, -18.4)
AD 86 (Rhinelander): PVI (0.0, -4.7, -13.2)
AD 87 (Merrill): PVI (1.9, -1.8, -7.4)

AD 88 (Ashland): PVI (4.5, 2.6, -4.8 )
AD 89 (Superior): PVI (9.1, 8.9, -1.9)
AD 90 (Osceola): PVI (-4.1, -6.2, -15.7)

Notice that the Pub swing was already underway from 2008 to 2012, with shifts of up to almost -5% in AD 86. But the shift for 2016 is really astounding with PVI jumps of -14% in AD 85, and almost -11% in the tradition Dem bastion of Superior. It defies common wisdom that AD 89 would support a Pub candidate, yet Trump won by 458 votes out of over 30K cast. This is one of the clearest indications I've seen of the effect of the Dem loss of the WWC and why Trump won WI.

I pulled the Eau Claire ward results for 2016, so here's the rest of the north:

AD 82 (Shawano): PVI (-1.0, -6.8, -17.4)
AD 83 (Wausaw): PVI (2.9, -1.9, -6.4)
AD 84 (Weston): PVI (-1.3, -7.9, -13.6)

AD 91 (Chippewa Falls): PVI (0.2, -2.7, -14.2)
AD 92 (Eau Claire - north): PVI (8.1, 5.9, 1.0)
AD 93 (Eau Claire - south): PVI (5.8, 2.9, 1.5)

AD 94 (Medford): PVI (-1.4, -8.7, -20.8 )
AD 95 (Black River Falls): PVI (8.1, 4.3, -8.3)
AD 96 (River Falls): PVI (1.4, -1.8, -9.6)

AD 97 (Rice Lake): PVI (-0.0, -1.6, -14.3)
AD 98 (Menomonee): PVI (0.5, -4.8, -8.7)
AD 99 (Hudson): PVI (-5.8, -8.1, -10.0)

Unlike Superior, Dem Eau Claire saw much less shift in its PVI. This probably reflects the different economies, with Eau Claire relying more on tech and the headquarters for Menards stores than Superior's role as a shipping port. The nearby AD 95 that had been Dem, saw a 13 point Pub swing in its PVI with an economy primarily in agriculture and tourism.

On the flip side, Pub Hudson saw only a small shift. It is primarily exurbs of the Twin Cities and like other suburban areas has more commuters and less WWC. Also like many other suburban college educated areas it had a strong 3rd party vote in 2016.

[jimrtex]

I find relatively few court decisions that would justify a range in excess of 10%. What you would have to show is that given your goals, there is no way to achieve them without exceeding 10%. That requires a clear statement of those goals, in a way that is not unduly arbitrary, so that one can test them to see if they can be achieved without the deviation.

At this point your goals seem to be:
1) minimize county chops by establishing a maximum number of regions;
2) minimize chops of munis within a region;
2) minimize the standard deviation of the population of the districts.

Feel free to correct this or add as necessary.

Abate v Mundt, Hawaii Supreme court decisions regarding canoe districts, Ohio decision regarding House district population (this was not limited to the single county districts), Jefferson v Tennant, Karcher v Daggett. There was the Wyoming decision by the SCOTUS where there were much larger deviations. And check out the deviations in Vermont (I assume they were litigated at some point).

Gerrymanderers prefer to disregard political subdivisions, and use population equality as a tool to force through their evil designs. Or they use the 10% rule to excuse their designs. The 10% total range rule is not an effective means of enforcing a good-faith effort at population equality. The Alabama legislature in 2000 did not act in good faith. The Arizona redistricting commission in 2010 did not act in good faith. The Florida courts in 2010 did not act in good faith to avoid dividing counties.

Here are a proposed set of rules:

Definitions:

Quota: Population base divided by total number of seats.
Normalized population: Population of a county divided by the quota.

Note: While I will display numbers as mixed decimal fractions rounded to three decimal points, or as percentages, any tests should be based on the use of rational numbers.

Entitlement: Number of districts equivalent to the population of the county (same as normalized population of the county).

Maximum whole districts: Entitlement truncated to whole number. This is the maximum number of districts with a population equal to the quota that may wholly be drawn in a county. In some instances we may be able to draw one more district. In those cases, the districts will be slightly undersized.

For counties with a population less than the quota, the maximum whole districts is zero.

Surplus: Entitlement minus maximum whole districts. The surplus will be in the interval [0, 1).

Apportionment region: One or more contiguous counties (excluding corner or near corner connectivity).

Characteristics of an apportionment region:

Total population: Sum of the normalized population.

Magnitude: Total population rounded to the nearest integer. This is the number of districts the region will be divided into. Magnitude must be one (1) or larger.

Average deviation: Total_population/Magnitude minus 1. This is expressed as a percentage of a quota. This is the relative deviation assuming that all districts in the region have equal population.

Absolute average deviation: Absolute value of average deviation.

If Magnitude is 1, then the absolute average deviation must be less than 7.5%.
If Magnitude is 2, then the absolute average deviation must be less than 6.0%
If Magnitude is 3 or more, then the absolute average deviation must be less than 5.0%.

If the relative deviation is larger than 5.0% it must be in relatively isolated cases.

Cumulative Surplus: Sum of the surpluses of all counties.

Whole number cumulative surplus: Cumulative surplus rounded to nearest integer.

Fraction cumulative surplus: Cumulative surplus minus whole number cumulative surplus, expressed as a percentage of a quota.

The fraction cumulative surplus must be in the range [-20%, 20%]. This is to ensure that a large error is not created in a region simply because we can spread it around among multiple counties. For example, this would prevent creation of single-county regions in Milwaukee and Waukesha.

Excess cumulative surplus. If the whole number cumulative surplus is greater than one, then the excess cumulative surplus is the whole number cumulative surplus minus one. This represents the excessive number of cuts (smaller counties being split, or surplus divided among multiple districts).

A putative division of an apportionment area should be done to determine the apparent number of smaller counties (less than a quota) that will be divided, as well as the division of a surplus between multiple districts. For example, a region containing Ozaukee, Milwaukee, and Racine will have more cuts than the mathematical minimum since a district that links all three counties would have to run the length of Milwaukee. The putative division need not take into account whether the division might require division of other political subdivisions.

Plans that would require fewer divided smaller counties are better than plans that would divide more smaller counties. The ideal is to divide no smaller counties.

Plans that would require fewer districts wholly in a larger county than the maximum whole districts are better than plans that would require more (e.g. a county with an entitlement of 1.423, is preferably divided into one district wholly within the county, and the surplus in multiple districts, is preferred to dividing the county into two large chunks).

Districts with fewer divided surpluses are generally better than those with more divided surplus.

RMS of Average Deviation is the square root of:
     the sum of the squares of the average deviation times the magnitudes
     divided by the total number of districts.

It should be less than 2.5%.

[jimrtex]

So you are allowing connections across lakes? Both Monona and Maple Bluff are surrounded by Madison except for frontage on a lake.

Monona has a land connection to Blooming Grove town, and I am treating Columbus-ized towns as being self-contiguous. Mud Lake is not a significant body of water in any case.

I believe that the original boundary between Westport and Madison towns did not follow the PLSS, and that Maple Grove is in the original area of Westport town on the north side of Lake Mendota. On a satellite view you can see the road on the township line between Westport and Burke on the eastern edge of Maple Grove.

This is distinct from the case of Shorewood Hills, which was clearly within Madison town, and there are no residual parts of Madison town near it. However, if there were a case where the population of Shorewood Hills prevented Madison from having a whole number of districts, I could see placing Shorewood Hills with Middleton.

Summary: I disagree with Monona. I can see your point with Maple Grove. When I make the division of Madison, I will take into account the possibility of adding Maple Grove.

[muon2]

So you are allowing connections across lakes? Both Monona and Maple Bluff are surrounded by Madison except for frontage on a lake.

Monona has a land connection to Blooming Grove town, and I am treating Columbus-ized towns as being self-contiguous. Mud Lake is not a significant body of water in any case.

I believe that the original boundary between Westport and Madison towns did not follow the PLSS, and that Maple Grove is in the original area of Westport town on the north side of Lake Mendota. On a satellite view you can see the road on the township line between Westport and Burke on the eastern edge of Maple Grove.

This is distinct from the case of Shorewood Hills, which was clearly within Madison town, and there are no residual parts of Madison town near it. However, if there were a case where the population of Shorewood Hills prevented Madison from having a whole number of districts, I could see placing Shorewood Hills with Middleton.

Summary: I disagree with Monona. I can see your point with Maple Grove. When I make the division of Madison, I will take into account the possibility of adding Maple Grove.

Mud Lake is wider than most rivers and there is no bridge. You can say they are contiguous, but to me that's like linking across a body like the Wisconsin river where there is no bridge. Your plan is legal, but much as you have your opinion about population deviation I strive for units that are actually connected.

[muon2]

Here are the 2016 updates for my Green Bay and Lake Winnebago regions.





AD 64 (Menasha): PVI (0.1, -4.0, -9.6)
AD 65 (Neenah): PVI (3.8, 2.3, -2.8 )
AD 66 (Oshkosh): PVI (2.7, 1.2, -2.9)

AD 67 (Ripon): PVI (-10.9, -14.4, -20.6)
AD 68 (Fond du Lac): PVI (-3.4, -3.7, -9.6)
AD 69 (Harrison): PVI (-3.8, -8.5, -14.3)

AD 70 (Sheboygan): PVI (4.9, 3.7, 0.7)
AD 71 (Plymouth): PVI (-11.9, -16.2, -18.1)
AD 72 (Manitowoc): PVI (2.2, -0.5, -9.4)

AD 73 (Grand Chute): PVI (-4.0, -9.8, -22.3)
AD 74 (Kaukauna): PVI (5.3, -1.3, -9.2)
AD 75 (Appleton): PVI (5.5, 3.0, 2.4)

AD 76 (Sturgeon Bay): PVI (3.5, -1.4, -8.9)
AD 77 (Green Bay - east): PVI (6.2, 3.5, -0.3)
AD 78 (Green Bay - west): PVI (7.1, 5.6, -0.5)

AD 79 (Suamico): PVI (-1.2, -7.1, -17.8 )
AD 80 (Howard): PVI (-2.6, -6.3, -8.8 )
AD 81 (De Pere): PVI (-2.1, -7.4, -8.9)

Again it's interesting to see where there were large shifts, and where they were more modest. There are both D (Appleton) and R (Howard and De Pere outside of Green Bay) areas with small shifts, and other areas with dramatic shifts well in excess of the state's swing.

[jimrtex]

So you are allowing connections across lakes? Both Monona and Maple Bluff are surrounded by Madison except for frontage on a lake.

Monona has a land connection to Blooming Grove town, and I am treating Columbus-ized towns as being self-contiguous. Mud Lake is not a significant body of water in any case.

I believe that the original boundary between Westport and Madison towns did not follow the PLSS, and that Maple Grove is in the original area of Westport town on the north side of Lake Mendota. On a satellite view you can see the road on the township line between Westport and Burke on the eastern edge of Maple Grove.

This is distinct from the case of Shorewood Hills, which was clearly within Madison town, and there are no residual parts of Madison town near it. However, if there were a case where the population of Shorewood Hills prevented Madison from having a whole number of districts, I could see placing Shorewood Hills with Middleton.

Summary: I disagree with Monona. I can see your point with Maple Grove. When I make the division of Madison, I will take into account the possibility of adding Maple Grove.

Mud Lake is wider than most rivers and there is no bridge. You can say they are contiguous, but to me that's like linking across a body like the Wisconsin river where there is no bridge. Your plan is legal, but much as you have your opinion about population deviation I strive for units that are actually connected.

You  can't get to the "part" of Madison that splits the two pieces of Blooming Grove from the state capitol without going through Blooming Grove.

And it is trivial to get between McFarland and Monona and spend less than a minute in Madison. If you asked for directions in McFarland to get to Monona, do you think someone would tell you that there was no way without going through Madison?