US House Redistricting: Minnesota

[Gass3268]

Any thoughts what the DFL would draw if they have total control in 2020?

[muon2]

Any thoughts what the DFL would draw if they have total control in 2020?

That probably depends on whether they want to violate the usual sensibilities by linking Mpls to outer suburbs. The public has not looked highly on those type of games in MN. Minnesota nice and all.

[Brittain33]

Any thoughts what the DFL would draw if they have total control in 2020?

That probably depends on whether they want to violate the usual sensibilities by linking Mpls to outer suburbs. The public has not looked highly on those type of games in MN. Minnesota nice and all.

No one except partisans cares enough about gerrymandering to take it out on the offending party, ever.

[Torie]

Oh nothing is absolute I suppose Muon2. But any system that allows a "discernible" amount of additional erosity for slightly more population equality, or forces a significant amount of additional erosity to avoid a chop, I think is flawed. We don't want to get a map from Dakota to Pine as being on the top of the heap absent highly unusual circumstances. I want erosity to have the most weight, chops next, with population at the bottom of the list.  How do you count population inequality? The sum of the amount of population of all the CD's that is above or below the perfect number?

I keep trying to focus on the practical, while I think you are trying to focus on the perfect, and what comes out leaves the madding crowd unhappy.

In your analysis of my map, were using that road chop thing again, that I can't understand very well, and I don't think you have sold anybody on? If we are going to generate more erosity to avoid road cuts, we are in trouble. (I did say I might be willing to count as a chop appending a county that has no state highway link to the balance of a CD.)

[ilikeverin]

Any thoughts what the DFL would draw if they have total control in 2020?

That probably depends on whether they want to violate the usual sensibilities by linking Mpls to outer suburbs. The public has not looked highly on those type of games in MN. Minnesota nice and all.

No one except partisans cares enough about gerrymandering to take it out on the offending party, ever.

Yes, but members of political parties are residents of their state, too.  Given Minnesotans' very limited experience with single-party control of government, anything could happen, including a coalition of Republicans and "Good Government" Democrats trying to force through a decent map.

[muon2]

In your analysis of my map, were using that road chop thing again, that I can't understand very well, and I don't think you have sold anybody on? If we are going to generate more erosity to avoid road cuts, we are in trouble. (I did say I might be willing to count as a chop appending a county that has no state highway link to the balance of a CD.)

My analysis of your map used a much simpler measure that usually gets the same ranking of two maps. Roads aren't considered at all for this measure.

Each county in the district counted the number of counties it bordered in other districts. Chopped counties count as separate counties for this measure. Add that up for all the counties in the district and you have the total I listed. Add up all the the districts and divide by two to get the total.

For example CD 1 in your map I counted 17 as follows:
Faribault (1): Martin
Blue Earth (3): Martin, Brown, Nicollet
Watonwan (4): Martin, Jackson, Cottonwood, Brown
Le Sueur (2): Nicollet, Sibley
Scott (5): Sibley, Carver, Hennepin W, Hennepin E, Dakota
Rice (1): Dakota
Goodhue (1): Dakota

The method is functionally like measuring the perimeter which is a good test for erosity. The use of roads is introduced to remove contiguous stretches like Watonwan-Jackson that have no convenient means for travel between. Its advantage over a perimeter measure is that it is immune to wiggly borders due to rivers and it gives a simple count that anyone readily compare when considering how chops affect erosity.

[Torie]

In your analysis of my map, were using that road chop thing again, that I can't understand very well, and I don't think you have sold anybody on? If we are going to generate more erosity to avoid road cuts, we are in trouble. (I did say I might be willing to count as a chop appending a county that has no state highway link to the balance of a CD.)

My analysis of your map used a much simpler measure that usually gets the same ranking of two maps. Roads aren't considered at all for this measure.

Each county in the district counted the number of counties it bordered in other districts. Chopped counties count as separate counties for this measure. Add that up for all the counties in the district and you have the total I listed. Add up all the the districts and divide by two to get the total.

For example CD 1 in your map I counted 17 as follows:
Faribault (1): Martin
Blue Earth (3): Martin, Brown, Nicollet
Watonwan (4): Martin, Jackson, Cottonwood, Brown
Le Sueur (2): Nicollet, Sibley
Scott (5): Sibley, Carver, Hennepin W, Hennepin E, Dakota
Rice (1): Dakota
Goodhue (1): Dakota

The method is functionally like measuring the perimeter which is a good test for erosity. The use of roads is introduced to remove contiguous stretches like Watonwan-Jackson that have no convenient means for travel between. Its advantage over a perimeter measure is that it is immune to wiggly borders due to rivers and it gives a simple count that anyone readily compare when considering how chops affect erosity.

Ah, Mike, below are two map options for the state of Atlasia. Each color represents a county. It has two CD's. Roads connect everything. Which map is more erose - the top one or the bottom one?

[minionofmidas]

Top obviously, not that bottom is great.

Your point being that it is technically possible to construct fantasy examples where the number of such county/district border matches is not just slightly but very much off?

[Torie]

Top obviously, not that bottom is great.

Your point being that it is technically possible to construct fantasy examples where the number of such county/district border matches is not just slightly but very much off?

Yes, probably for Midwest grid county arrangements, of roughly the same size, and lined up, Mike's little system is a pretty good proxy. But it can go seriously wrong, with odder shaped counties (the specific problem being you get screwed if a big county borders a bunch of little ones in another CD, as opposed to another big county, even though it may have no impact on erosity whatsoever, except in Mike's world ). I would rather avoid proxies, and go for the real deal, which relates to perimeter lengths, and how much of a CD fits into a perfect square. Fitting into circles isn't great, because God made most counties into rectangles, not circles.

[minionofmidas]

The results should probably hold up anywhere east of the Great American Desert.

[muon2]

Top obviously, not that bottom is great.

Your point being that it is technically possible to construct fantasy examples where the number of such county/district border matches is not just slightly but very much off?

Yes, probably for Midwest grid county arrangements, of roughly the same size, and lined up, Mike's little system is a pretty good proxy. But it can go seriously wrong, with odder shaped counties (the specific problem being you get screwed if a big county borders a bunch of little ones in another CD, as opposed to another big county, even though it may have no impact on erosity whatsoever, except in Mike's world ). I would rather avoid proxies, and go for the real deal, which relates to perimeter lengths, and how much of a CD fits into a perfect square. Fitting into circles isn't great, because God made most counties into rectangles, not circles.

What you ask for already exists. It is the statutory system for IA and is designed to make districts as square as possible within the constraint of population equality and maintaining political subdivisions. I've seen no effective suggestions of a way to improve it without going to the circle-based methods or even more complex algorithms that combine circles and blockwise density values. That's not to say it doesn't have detractors.

This is the relevant statute:

You must be logged in to read this quote.

Note that there are two tests involved and both are relevant with neither given explicit priority though I would note that IA plans seem to favor point 4a when the two are in confict. That would be consistent with the goal of getting square districts. It also would be the priority if a state had more counties with irregular boundaries as we see comparing MN to IA. By the way hexagonal in this context is the shape that results from three rows of staggered squares.

Let's apply the IA regimen to your examples. I will assume that the brown L-shaped county and pale strip county have the same population so that swapping them does not change any population inequality measures. I will also set your grid to 1 mile, so that the smallest county is 1 mile by 1 mile. I'll call the district that includes the NE-most county CD1 and the other CD2.

First plan:
test 4a: CD1 NS=4 mi, EW=7 mi, diff=3 mi; CD2 NS=6 mi, EW=10 mi, diff=4 mi; total=7 mi.
test 4b: CD1 perimeter = 24 mi; CD2 perimeter = 40 mi; total = 64 mi.

Second plan:
test 4a: CD1 NS=6 mi, EW=10 mi, diff=4 mi; CD2 NS=3 mi, EW=8 mi, diff=5 mi; total=9 mi.
test 4b: CD1 perimeter = 32 mi; CD2 perimeter = 22 mi; total = 54 mi.

Using the gold standard of square box redistricting from IA I could as easily choose the first plan as the second, and often would. I suspect that is not the answer you want from a system. This is a problem with synthetic examples - they tend to bollix up multiple systems. Any of the systems we discuss for redistricting is a proxy designed to use pure geography to get at communities of interest, with the idea that convenience of distance relative to all points in a district will reflect that goal. County integrity and erosity both fall into that category.

[Torie]

I will try to parse through you post Mike, but in the meantime, it is true is it not, that your system counts the more erose plan vis a vis anyone's common sense, as in fact considerably less erose?  We need to peel back the layers on the onion one at a time.

[jimrtex]

I wrote a really long explanation which was lost.

Anyhow, the colored areas are the central counties of the larger metropolitan areas (more than 10% of a congressional district), exluding the Minnesota portions of the Fargo, Grand Forks, and La Crosse metro areas.  Central counties are defined based on the presence of urban areas, densely, continuously populated areas.

Summary: Central counties are a better measure of metropolitan areas for redistricting purposes, and provide a measure for preferred shedding counties, the percentage of the county population within the core urbanized area:

Ramsey 100%
Hennepin 98%
Dakota 89%
Anoka 85%
Washington 78%
Scott 68%
Carver 61%
Wright 27%
Sherburne 22%

Clearly the latter two are peripheral with the Urbanized Area reaching the two along I-94 and US 52 up the Mississippi rather than more general spillover.

The outlying counties of Le Sueur, Sibley, Mille Lacs, Isanti, Chisago, St Croix (Wisc.), and Pierce (Wisc.) are included on the basis of commuting patterns, which may simply reflect the relatively dearth of non-agricultural jobs in rural counties.  And while Sibley and Mille Lacs are included, Rice is not, because Faribault does provide a source of jobs.

[jimrtex]

This map confirms that a 4-3 map should be drawn.

Possible configurations:

A) 4-1-2

1) St. Cloud-Outer Metro North, Chisago, Isanti, Benton, Stearns, Benton, Sherburne, and Wright, plus areas to the north and west.  The metro area is shifted south to include Goodhue, Rice, and Le Sueur.
2) Duluth-Grand Forks- Fargo, and coming way south past the the Minnesota River.
3) Southern Minnesota, La Crosse to Sioux Falls, anchored in Rochester and Mankato.

B) 4-3

1) Duluth-Northern Exurbs-St.Cloud
2) Western Minnesota - Manitoba to Iowa, with a somewhat irregular border.
3) Southeastern Minnesota.

C) 4-3

1) Duluth-Northern Exutbs and points west.
2) St. Cloud to southwest corner.
3) Southeastern Minnesota.


ps Can anyone get the minor civil division estimates for 2011 or 2012 for Lake of the Woods, Hennepin, and Koochiching counties.  I noticed Lake of the Woods first, and thought maybe there was something odd about the county, but then I found the other two as well.  Either the Census Bureau doesn't have the data, or their is some weird interaction  with my browser.


It happens that it may be possible to make districts largely out of whole counties in the Metro area.   Ramsey+Washington, Hennepin+Anoka (2), with northern and western Hennepin with Anoka.   Dakota+Scott+Carver+Wright.

If a Minneapolis-St.Paul district is created, then a chop of Ramsey would be needed: Anoka+Washington+Northern Ramsey.  The district wholly in Hennepin then would be the suburbs.

[muon2]

I will try to parse through you post Mike, but in the meantime, it is true is it not, that your system counts the more erose plan vis a vis anyone's common sense, as in fact considerably less erose?  We need to peel back the layers on the onion one at a time.

My post is to point out that what your eye sees as more erose is not necessarily so when measured by known algorithms designed to promote square districts. That my novel technique (from an idea by jimrtex) also classifies them in a similar way does not indicate a weakness in my system. It points out what is well known - it is hard to bias against C-shaped districts in favor of L-shaped ones without creating all sorts of other problems.

The thought behind my method is anything but pursuing the perfect. It is in recognition of the weaknesses of more detailed algorithms that I went for a KISS approach. I'm willing to preserve that same weakness in the name of a simpler method. It's one that anyone with Google maps can check. Your examples illustrate how simple the method is. The only part that I need to tune in my method is how to measure erosity in chopped counties. That's a task I'm actively pursuing.

[Torie]

I will try to parse through you post Mike, but in the meantime, it is true is it not, that your system counts the more erose plan vis a vis anyone's common sense, as in fact considerably less erose?  We need to peel back the layers on the onion one at a time.

My post is to point out that what your eye sees as more erose is not necessarily so when measured by known algorithms designed to promote square districts. That my novel technique (from an idea by jimrtex) also classifies them in a similar way does not indicate a weakness in my system. It points out what is well known - it is hard to bias against C-shaped districts in favor of L-shaped ones without creating all sorts of other problems.

The thought behind my method is anything but pursuing the perfect. It is in recognition of the weaknesses of more detailed algorithms that I went for a KISS approach. I'm willing to preserve that same weakness in the name of a simpler method. It's one that anyone with Google maps can check. Your examples illustrate how simple the method is. The only part that I need to tune in my method is how to measure chopped counties. That's a task I'm actively pursuing.

I guess I will accept that deflection of my question as an admission. Thank you. In any event, it is not just my eye, it's almost everyone's eye (you, you math/physicist person happiest when caressing the supercollider like a pedigree dog,  just see the beauty of the algorithm as overshadowing the bestial veneer, when maps are all about two dimensional veneers. Can you detail in a clear way, just why any cure to get the box to match the eye, is a fool's errand, with the cure worse than the disease?

[muon2]

I will try to parse through you post Mike, but in the meantime, it is true is it not, that your system counts the more erose plan vis a vis anyone's common sense, as in fact considerably less erose?  We need to peel back the layers on the onion one at a time.

My post is to point out that what your eye sees as more erose is not necessarily so when measured by known algorithms designed to promote square districts. That my novel technique (from an idea by jimrtex) also classifies them in a similar way does not indicate a weakness in my system. It points out what is well known - it is hard to bias against C-shaped districts in favor of L-shaped ones without creating all sorts of other problems.

The thought behind my method is anything but pursuing the perfect. It is in recognition of the weaknesses of more detailed algorithms that I went for a KISS approach. I'm willing to preserve that same weakness in the name of a simpler method. It's one that anyone with Google maps can check. Your examples illustrate how simple the method is. The only part that I need to tune in my method is how to measure chopped counties. That's a task I'm actively pursuing.

I guess I will accept that deflection of my question as an admission. Thank you. In any event, it is not just my eye, it's almost everyone's eye (you, you math/physicist person happiest when caressing the supercollider like a pedigree dog,  just see the beauty of the algorithm as overshadowing the bestial veneer, when maps are all about two dimensional veneers. Can you detail in a clear way, just why any cure to get the box to match the eye, is a fool's errand, with the cure worse than the disease?

The simple example is C vs L. Consider a 3 by 3 grid of square counties and denote them A,B,C in row 1, D,E,F in row 2, and G,H,I in row three. To the average eye, a district that contains all but counties E and F looks suspicious. It's a C-shape and one wonders why the peninsula is sticks in in such a way, could it be gerrymandering? IA even has a special rule to deal with that type of shape and bans it since C and I are entirely separated by F which would be in another district, but that doesn't help in your given example. Note that this hypothetical district has 7/9 of the smallest square that contains the district.

Now picture a district that has A,B,C,D, and G. It forms an L shape that seems less erose than my first example, but it only covers 5/9 of the smallest enclosing square. That means it won't score as less erose using a covering square as the system. On a N-S/E-W measure it is only equal to the C-shape I first described, so that won't help.

So why not look at the perimeter alone. The C-shape has a perimeter of 16 and the L-shape only 12. That can work for those two districts, but consider a chop of one county into a C-shape. The dimensions are reduced by a factor of 3 so its perimeter looks good compared to the L-shape. Perimeter measurements alone tend to favor good looking large districts at the expenses of potentially strange-looking small districts.

There are more complex measurements involving the ratio of the square of the perimeter to the area, and they work well when everything is square blocks like IA. Once you go to irregular counties like those along the Minnesota and Mississippi rivers in MN, or most of the counties in a state like TN then this method starts to penalize districts that follow the natural lines that form the counties. I was able to game that rule in the 2009 OH competition by avoiding districts that followed those natural lines. If natural boundaries should have meaning then this is not a good solution either.

My conclusion was that by simplifying the measurement of erosity it actually becomes harder to game the system. It helps prevent gaming simply by providing fewer knobs to play with.

[muon2]

This map confirms that a 4-3 map should be drawn.

Possible configurations:

A) 4-1-2

1) St. Cloud-Outer Metro North, Chisago, Isanti, Benton, Stearns, Benton, Sherburne, and Wright, plus areas to the north and west.  The metro area is shifted south to include Goodhue, Rice, and Le Sueur.
2) Duluth-Grand Forks- Fargo, and coming way south past the the Minnesota River.
3) Southern Minnesota, La Crosse to Sioux Falls, anchored in Rochester and Mankato.

B) 4-3

1) Duluth-Northern Exurbs-St.Cloud
2) Western Minnesota - Manitoba to Iowa, with a somewhat irregular border.
3) Southeastern Minnesota.

C) 4-3

1) Duluth-Northern Exutbs and points west.
2) St. Cloud to southwest corner.
3) Southeastern Minnesota.

I agree on the general idea, but my plan seems a bit of a hybrid. Ignoring Isanti as a tool to equalize population, it looks like I have something like A, but a bit of C where I attach some of the exurban north metro to the Duluth district. That keeps the Duluth district in the north and St Cloud away from the SE. Here would be my revised A:

D) 4-1-2

1) St. Cloud-Outer Metro North West: Stearns, Benton, Sherburne, and Wright, plus areas to the west including Fargo.  The metro area is shifted south to include Goodhue, Rice, and Le Sueur.
2) Northern Minnesota: Duluth to Grand Forks including the northern exurbs.
3) Southern Minnesota: La Crosse to Sioux Falls, anchored in Rochester and Mankato.

[jimrtex]

Possible configurations:

A) 4-1-2

1) St. Cloud-Outer Metro North, Chisago, Isanti, Benton, Stearns, Benton, Sherburne, and Wright, plus areas to the north and west.  The metro area is shifted south to include Goodhue, Rice, and Le Sueur.
2) Duluth-Grand Forks- Fargo, and coming way south past the the Minnesota River.
3) Southern Minnesota, La Crosse to Sioux Falls, anchored in Rochester and Mankato.

I rather like this.   The 4 metro districts total 4.007.  If you want to equalize a bit more, trim a tiny bit off Anoka (say, Linwood).

You have the cut of Hennepin, and then have to take about 25,000 from the south. 

Some possibilities include

a) Chanhassen, though I don't like splitting it from Chaska.

b) South St. Paul or West St. Paul, with a shift through Ramsey or Washington.

c) Hasting, with a shift through Ramsey or Washington.  I think I prefer Hastings since it is somewhat isolated from most of Dakota, and does cross the Mississippi.

[jimrtex]

D) 4-1-2

1) St. Cloud-Outer Metro North West: Stearns, Benton, Sherburne, and Wright, plus areas to the west including Fargo.  The metro area is shifted south to include Goodhue, Rice, and Le Sueur.
2) Northern Minnesota: Duluth to Grand Forks including the northern exurbs.
3) Southern Minnesota: La Crosse to Sioux Falls, anchored in Rochester and Mankato.

I think that the split of the Red River will be a hard sale.

[jimrtex]

Possible configurations:

B) 4-3

1) Duluth-Northern Exurbs-St.Cloud
2) Western Minnesota - Manitoba to Iowa, with a somewhat irregular border.
3) Southeastern Minnesota.

This preserves the northern Red River/Iron Range-Great Lakes split.  I think a swap of St.Cloud for Brainerd and Bemidji might be better.

McLeod and LeSueur were swapped to give the Metro districts 4.001.  Reversing the swap, the Metro districts need about 7000 persons.   An alternative would be to take a little bit from Sherburne, Isanti, or Chisago.

[jimrtex]

Possible configurations:

B) 4-3

1) Duluth-Northern Exurbs-St.Cloud
2) Western Minnesota - Manitoba to Iowa, with a somewhat irregular border.
3) Southeastern Minnesota.

This shifts LeSueur to the south, and makes creates a clear Manitoba-Iowa district.  The metro area will need about 27,000 from Sherburne, Isanti, or Chisago (likely the SE corner of Sherburne).

[jimrtex]

Possible configurations:

B) 4-3

1) Duluth-Northern Exurbs-St.Cloud
2) Western Minnesota - Manitoba to Iowa, with a somewhat irregular border.
3) Southeastern Minnesota.

This version exchanges St.Cloud for Brainerd and Bemidji.  Like the previous version this will require about 27,000 persons shifted from Sherburne to the metro area.

[jimrtex]

Possible configurations:

C) 4-3

1) Duluth-Northern Exutbs and points west.
2) St. Cloud to southwest corner.
3) Southeastern Minnesota.

The previous version, Plan B2, shows that a St.Cloud-South West district stretches North to the Canadian border.  To have a northern district, it has to be pushed out of the Metro exurbs.  This map shifts Sherburne to the St.Cloud-Southwest, but the northern district divides Fargo and Grand Forks.



This version takes the St.Cloud-Southwest district across the northern exurbs, and creates a northern district north of Duluth-Brainerd-Fargo.   It is the same as in Plan A.  About 27,000 persons will be shifted from Sherburne, Isanti, or Chisago to the metro districts.



This shifts Wright to the St.Cloud-Southwest district.  I think this is the best configuration for such a district, avoiding going north to Fargo, or east to Wisconsin.  To compensate for the the loss of Wright, the south metro district will need about 135,000 from either Hennepin or Washington, which forces the other districts to take in the northern exurbs.

If Ramsey-Washington is maintained, then Hennepin will need to be double chopped.

The alternative which is implied by this map is to create the Minneapolis-St.Paul district.   That will require a Ramsey split, and likely Washington and Anoka splitts.

[muon2]

Possible configurations:

B) 4-3

1) Duluth-Northern Exurbs-St.Cloud
2) Western Minnesota - Manitoba to Iowa, with a somewhat irregular border.
3) Southeastern Minnesota.

This shifts LeSueur to the south, and makes creates a clear Manitoba-Iowa district.  The metro area will need about 27,000 from Sherburne, Isanti, or Chisago (likely the SE corner of Sherburne).

If we are looking at these with an eye towards 2020, the growth will be predominantly in the TC metro. My projections are that Hennepin+Anoka will be about 43K larger than two CDs and Ramsey+Washington will be about 20K larger than a CD. The four counties that wrap from Wright to Dakota will be about 16K larger than a CD. Those metro counties will have to shed population to the rest of the state - most likely by moving Wright to St Cloud's district and running the south suburbs to Rice+.