If every state had NC-style clusters, what would they end up like?

[jimrtex]

What's an NC cluster?

North Carolina minimizes county splits on state legislative level by having groups of counties in which a given number of seats must be nested. These are typically as small as possible.

The North Carolina constitution forbids division of counties. It does permit multi-member districts and multi-county districts.

In larger counties, VRA requirements likely require single-member districts. Formerly (before 2010) VRA districts were drawn in rural areas. There are very few counties with a majority black population, so these districts tended to snake across the landscape trying to connect areas with a majority black population, perhaps dividing small cities based on where the black population lives.

Over time the rural black population has declined, moving first to Washington, Philadelphia, New York, and other northern cities, and now to cities such as Atlanta, Charlotte, and Raleigh where there are better economic opportunities. The overall increase in the total population has required larger legislative districts. If you have stretched across multiple counties to create a black-majority district, and you have to add population, you may be unable to maintain a black majority. This has particularly happened in southeastern North Carolina. At some point you may be assigning persons to districts based on race.

In rural areas of the northeast where there is a substantial black population, you might not have to draw snakes, because you may be packing. If a county is 50% black, the areas where blacks predominate may be 60% or 70% black. In addition, the Gingles condition that whites vote as bloc to prevent election of the black candidate of choice may no  longer hold. With black turnout as high as or exceeding white turnout in many areas, there might be a relatively small cross-over vote needed.

Through 2010, and the initial 2010 redistricting, the VRA districts were drawn first, and then county clusters were drawn around them. After mid-decade litigation, the snakes were eliminated and pure county clustering was used.

The reason for county clustering is that it minimizes county cutting which violates the NC Constitution. Unlike in Texas, where single-member districts in large counties is considered to be a constitutional manner regulation, it is considered a violation of the NC Constitution. Nevertheless, if you draw a whole number of districts in Wake or Mecklenburg or other large counties you avoid an additional split of a district that crosses a county line.

The NC Supreme Court has also ruled that 2-member districts violate equal protection (this might be under the NC Constitution rather than the US Constitution) because it lets some voters vote for two representatives rather than just one, even if there twice as many of them. You might have previously been able to draw a group of counties that had a population equal to two representatives and elected them at large, this is no longer possible. Instead you are going to split one of the counties.

Generally more clusters means fewer county cuts. If you have a county cluster entitled to three representatives, and you can split it on county lines such that one cluster has two representatives, and the other one, and have eliminated a cut.

If clusters have fewer counties, this will tend to produce more of them. There is a belief in North Carolina that you should draw the one county clusters first, then the two-county clusters, and so on. This might not produce the minimum number of clusters, but will generally do a fairly good job.

However it may force fewer clusters and greater deviation. Whenever you set off one cluster you must also ensure that any remaining contiguous groups of counties can be apportioned a whole number of representatives.

This may increase the deviation and require more county cuts. Imagine that you apportion five representatives to an area entitled to 4.85 representatives. This creates an unreasonable -3.0% deviation. But it also forces the 0.15 representatives to be placed in other districts. And what if that area is entitled to 3.15 representatives. Each district must have a population of exactly 1.0500000 quotas. You may be forcing another cut. If the area was entitled to 3.08 representatives you might have been able to divide it on a county line.

This happened in North Carolina in the mid-2010 redistricting. If you look at the area south of Raleigh you will find such districts where little nibbles of counties were bitten off to get exact populations just as offensive as when doing congressional redistricting.

If I were scoring a North Carolina clustering plan, I would score it based on implied county cuts and lowest standard deviation.

Or better based on Texas, I would only penalize division of smaller counties - less than a quota. Ties would be broken based on lowest standard deviation.

[jimrtex]

What's an NC cluster?

North Carolina minimizes county splits on state legislative level by having groups of counties in which a given number of seats must be nested. These are typically as small as possible.

The North Carolina constitution forbids division of counties. It does permit multi-member districts and multi-county districts.

In larger counties, VRA requirements likely require single-member districts. Formerly (before 2010) VRA districts were drawn in rural areas. There are very few counties with a majority black population, so these districts tended to snake across the landscape trying to connect areas with a majority black population, perhaps dividing small cities based on where the black population lives.

Over time the rural black population has declined, moving first to Washington, Philadelphia, New York, and other northern cities, and now to cities such as Atlanta, Charlotte, and Raleigh where there are better economic opportunities. The overall increase in the total population has required larger legislative districts. If you have stretched across multiple counties to create a black-majority district, and you have to add population, you may be unable to maintain a black majority. This has particularly happened in southeastern North Carolina. At some point you may be assigning persons to districts based on race.

In rural areas of the northeast where there is a substantial black population, you might not have to draw snakes, because you may be packing. If a county is 50% black, the areas where blacks predominate may be 60% or 70% black. In addition, the Gingles condition that whites vote as bloc to prevent election of the black candidate of choice may no  longer hold. With black turnout as high as or exceeding white turnout in many areas, there might be a relatively small cross-over vote needed.

Through 2010, and the initial 2010 redistricting, the VRA districts were drawn first, and then county clusters were drawn around them. After mid-decade litigation, the snakes were eliminated and pure county clustering was used.

The reason for county clustering is that it minimizes county cutting which violates the NC Constitution. Unlike in Texas, where single-member districts in large counties is considered to be a constitutional manner regulation, it is considered a violation of the NC Constitution. Nevertheless, if you draw a whole number of districts in Wake or Mecklenburg or other large counties you avoid an additional split of a district that crosses a county line.

The NC Supreme Court has also ruled that 2-member districts violate equal protection (this might be under the NC Constitution rather than the US Constitution) because it lets some voters vote for two representatives rather than just one, even if there twice as many of them. You might have previously been able to draw a group of counties that had a population equal to two representatives and elected them at large, this is no longer possible. Instead you are going to split one of the counties.

Generally more clusters means fewer county cuts. If you have a county cluster entitled to three representatives, and you can split it on county lines such that one cluster has two representatives, and the other one, and have eliminated a cut.

If clusters have fewer counties, this will tend to produce more of them. There is a belief in North Carolina that you should draw the one county clusters first, then the two-county clusters, and so on. This might not produce the minimum number of clusters, but will generally do a fairly good job.

However it may force fewer clusters and greater deviation. Whenever you set off one cluster you must also ensure that any remaining contiguous groups of counties can be apportioned a whole number of representatives.

This may increase the deviation and require more county cuts. Imagine that you apportion five representatives to an area entitled to 4.85 representatives. This creates an unreasonable -3.0% deviation. But it also forces the 0.15 representatives to be placed in other districts. And what if that area is entitled to 3.15 representatives. Each district must have a population of exactly 1.0500000 quotas. You may be forcing another cut. If the area was entitled to 3.08 representatives you might have been able to divide it on a county line.

This happened in North Carolina in the mid-2010 redistricting. If you look at the area south of Raleigh you will find such districts where little nibbles of counties were bitten off to get exact populations just as offensive as when doing congressional redistricting.

If I were scoring a North Carolina clustering plan, I would score it based on implied county cuts and lowest standard deviation.

Or better based on Texas, I would only penalize division of smaller counties - less than a quota. Ties would be broken based on lowest standard deviation.

I don't want to clip your post, the whole thing does a great job summarizing the history of how the clusters come about, but the examples you give in the bolded paragraph are actually reasonable in deviation under North Carolina's current rules. As long as population does not deviate by more than ±5% off of the ideal district size, it is allowed. So a cluster entitled to 4.85 representatives getting 5 or a cluster entitled to 3.15 representatives getting 3 is actually within that acceptable range. You could even lose a tenth of an entitled delegate in the first example and it would still be fine.

This does create some interesting potentials for states with large counties and lots of reps. Once you start getting up into the double-digit level for entitled representatives, the population deviations start to overlap. A county or county-cluster that is given 10 delegates might be due anywhere from 9.5 to 10.5 based on its population. A district given 11 delegates is due between 10.45 and 11.55 delegates. That overlap on the upper- and lower-ends makes for some variation that depends on the smaller counties and the clusters they make up. So even though county-clusters are determined first, delegates are not necessarily determined until after the clusters are made because potentially larger counties might be given or need to shed a district to optimize down the line.

In North Carolina, our two largest counties are Mechlenburg and Wake, and both received 13 delegates, and both could've only received 13 delegates because they were only within that range. However, in other states where single counties are more dominant, this might come into play. For example, Cook county, Illinois would be entitled to an ideal of ~48.59 districts, but could actually receive between 47 and 51 of the State House's 118 districts depending on what creates optimal later clusters if they were operating under North Carolina's own standards.

I didn't describe the problem with the cluster that has a population equivalent to 3.15 representatives very well.

In that case the three districts had to have populations of exactly 1.05 quotas. Not only are you going to have to cross county lines, you are going to have to cross any other kind of sensible boundary just to keep all three districts within 5%.

IIRC in the case of North Carolina it was not possible to draw districts cutting only two counties in the cluster with magnitude 3 or they would have been awful.

Or consider a larger area entitled to 5.20 delegates. You might not be able to split it into two clusters because you have to distribute the excess between the clusters. If you have a an cluster with a perfect 2.0 delegates, the other is an excessive 3.20.

The problem in NC is not having cluster, but the particular algorithm they use to identify them (unless they have abandoned that).

I think in your Illinois case you may have an equal protection violation since you will be overpopulating every Cook County district, and underpopulating every downstate district (for this purpose, Lake County is downstate), or vice versa.

[jimrtex]

Some 8-cluster plans for a 35-member Maine Senate I've come up with:

Androscoggin (3 districts; 2.8552 or 3×0.9517 quotas; −4.83%)
Aroostook-Washington-Hancock-Penobscot-Piscataquis (8 districts; 8.2898 or 8×1.0362 quotas; +3.62%)
Cumberland (8 districts; 7.7861 or 8×0.9733 quotas; −2.67%)
Franklin-Somerset (2 districts; 2.0535 or 2×1.0268 quotas; +2.68%)
Kennebec-Lincoln-Sagadahoc (5 districts; 5.0245 or 5×1.0049 quotas; +0.49%)
Waldo (1 district; 1.0175 quotas; +1.75%)
Knox (1 district; 1.0432 quotas; +4.32%)
Oxford-York (7 districts; 6.9300 or 7×0.9900 quotas; −1.00%)

Androscoggin (3 districts; 2.8552 or 3×0.9517 quotas; −4.83%)
Aroostook-Washington-Hancock (4 districts; 3.9481 or 4×0.9870 quotas; −1.30%)
Cumberland-York (13 districts; 13.2318 or 13×1.0178 quotas; +1.78%)
Kennebec-Lincoln-Sagadahoc (5 districts; 5.0245 or 5×1.0049 quotas; +0.49%)
Waldo (1 district; 1.0175 quotas; +1.75%)
Knox (1 district; 1.0432 quotas; +4.32%)
Oxford-Franklin-Somerset-Piscataquis (4 districts; 3.9695 or 4×0.9924 quotas; −0.76%)
Penobscot (4 districts; 3.9101 or 4×0.9775 quotas; −2.25%)

What I thought I had for a 9-cluster plan:

Androscoggin (3 districts; 2.8552 or 3×0.9517 quotas; −4.83%)
Aroostook-Washington-Hancock (4 districts; 3.9481 or 4×0.9870 quotas; −1.30%)
Cumberland (8 districts; 7.7861 or 8×0.9733 quotas; −2.67%)
Franklin-Somerset (2 districts; 2.0535 or 2×1.0268 quotas; +2.68%)
Kennebec-Lincoln-Sagadahoc (5 districts; 5.0245 or 5×1.0049 quotas; +0.49%)
Waldo (1 district; 1.0175 quotas; +1.75%)
Knox (1 district; 1.0432 quotas; +4.32%)
Oxford-York (7 districts; 6.9300 or 7×0.9900 quotas; −1.00%)
Penobscot (4 districts; 3.9101 or 4×0.9775 quotas; −2.25%)
Piscataquis (0 districts; 0.4316 or 0×∞ quotas; +∞%)

How many are Texas-style where you apportion a whole number of districts to large counties, and one district for any surplus or smaller counties. The goal is to avoid dividing counties with a population smaller than a quota.

[jimrtex]

MN, with nestings
quota for Senate is 85,172; for House, 42,586.
Cook+Lake+Carlton+St. Louis: 3 Senate, 6 House
Pine+Kanabec+Aitkin+Mille Lacs: 1 Senate, 2 House
Itasca+Cass+Wadena: 1 Senate, 2 House
Koochiching+Lake of the Woods+Hubbard+Beltrami: 1 Senate, 2 House
Roseau+Kittson+Marshal+Pennington+Red Lake+Polk+Norman: 1 Senate, 2 House
Clearwater+Mahnomen+Becker+Clay+Otter Tail: 2 Senate, 4 House
Crow Wing+Morrison+Todd+Douglas+Grant: 2 Senate, 4 House
Pope+Stevens+Wilkin+Traverse+Big Stone+Lac qui Parle+Yellow Medicine+Lyon+Lincoln: 1 Senate, 2 House
Benton+Stearns+Wright: 4 Senate, 8 House
Isanti+Sherburne+Anoke: 6 Senate, 12 House
Chisago+Washington+Ramsey: 10 Senate, 20 House
Kandiyohi+Swift+Chippewa+Renville: 1 Senate, 2 House
Carver+McLeod+Meeker: 2 Senate, 4 House
Hennipen: 15 Senate, 30 House
Dakota: 5 Senate, 10 House
Scott+Sibley: 2 Senate, 4 House
Redwood+Brown+Watonwan+Cottonwood+Murray+Pipestone: 1 Senate, 2 House
Rock+Nobles+Jackson+Martin+Faribault+Freeborn+Steele+Waseca: 2 Senate, 4 House
Blue Earth+Nicollet+Le Sueur+Rice+Goodhue: 3 Senate, 6 House
Olmsted: 2 Senate, 4 House
Wabasha+Winona+Houston+Fillmore+Dodge+Mower: 2 Senate, 4 House
https://davesredistricting.org/join/b44a3f90-5be7-4db6-800e-6918a01d3a9f
35 SDs and 70 HDs in Biden-voting clusters, out of 67/134 respectively.

I would score the Senate plan on (1) number of small county cuts; (2) modest large county cuts; and (3) population equality. There could conceivably be a bonus if a house district can be divided at a county line (e.g. a 2 district senate cluster of smaller counties would require a county split for two house districts, but one or both of the other two districts might be made of whole counties.

Cook+Lake+Carlton+St. Louis, might be scored as one type (2) cut IF you could plausibly locate two whole districts in Saint Louis. I don't know whether you can connect the other three counties by a corridor looping around the Iron Range, or squeezing between the Iron Range and Duluth. It might be preferable to only have one whole district in Saint Louis and one district that is partially in the county along with Carlton, and the other in the county plus the Arrowhead. In that case, it might be OK to accept a larger penalty that other plans can not defeat.

I would count the failure to have two whole districts in Saint Louis as one type (1) cut and two type (2) cuts.

Clearwater+Mahnomen+Becker+Clay+Otter Tail: This is one type (1) cut. I would want a demonstration that the cut (presumably Becker) is plausible in not requiring a township split. You might be able to move the cut to Clay or Otter Tail. It would then come down to communities of interest - but would not score differently so long as only one county was cut.

Crow Wing+Morrison+Todd+Douglas+Grant: One type (1) cut. Demonstrate cut of Morrison is plausible.

Pope+Stevens+Wilkin+Traverse+Big Stone+Lac qui Parle+Yellow Medicine+Lyon+Lincoln. Possible bonus if House districts don't require county cut (e.g. Lyon-Lincoln-Yellow Medicine, other six in the other so long as it is within 5%.

Benton+Stearns+Wright: Two type (2) cuts for Stearns and Wright. I would also require a plausible demonstration of the leftovers district. You might end up with a type (1) cut in Stearns.

Isanti+Sherburne+Anoke: Two type (2) cuts.

Chisago+Washington+Ramsey: The creation of 6 districts in Ramsey and 3 in Washington is fine. Demonstrate the Ramsey-Washington-Chisago district. In addition, it might not be acceptable to treat Washington as having a surplus, since the total surplus for the area is greater than the surplus is greater than for Washington alone.

What happens with Washington and Chisago-Pine? Or perhaps Ramsey-Anoka-Washington with one district at the junction of the three counties.

Kandiyohi+Swift+Chippewa+Renville: This has a -5.03% deviation. I am fine with this so long as overall deviation is OK.

Carver+McLeod+Meeker: One type (2) split.

Dakota - a court would reasonably ask why Washington had a surplus but Dakota did not.

Scott+Sibley - one type (2) split.

Rock+Nobles+Jackson+Martin+Faribault+Freeborn+Steele+Waseca: A type (1) split. Waseca-Steele-Freeborn is a possible senate district. If you drew the map from north to south, you may have trapped yourself in having to get districts in the south barely in range.

Blue Earth+Nicollet+Le Sueur+Rice+Goodhue Two type (1) splits assuming the cuts are in Nicollet/Blue Earth and Rice.

Wabasha+Winona+Houston+Fillmore+Dodge+Mower: One type (1) split.

Wabasha+Winona+Houston+Fillmore+Dodge+Mower: One type (1) split. I might go with Wabasha+Winona+Houston and Fillmore+Dodge+Mower and go with an exceptional deviation. Alternatively place the three river counties with Olmsted, which reduces the deviation, while adding a type (2) split.

Standard deviation is 2.40 which would be high by Texas standards, but perhaps not in Minnesota due to the relative high number of districts relative to number of counties.

The Minnesota Constitution does specify the number of Senate districts, though it does require nesting, though not necessarily 2:1. In the early history the nesting ratio might vary throughout the state, but that would not comply with modern standards of equal population. A similar constitution in Washington, Arizona, and North Dakota has been interpreted as permitting multi-member House districts, either elected by position as in Washington, or as two-member districts like in Arizona and North Dakota.

Senate districts are explicitly single-member. Since there is not the same requirement for House districts it is implicit that the framers wanted to make that optional (rather than be carelessly inconsistent in their language).

It would be interesting to see if a better Minnesota map could be drawn by a slight variation in the Senate size say in the range of 64 to 70.

More radically might be 1:5 ratio with 33 or 34 senators and 165 or 170 representatives to maintain the overall size of the legislature, and to elect the representatives by STV. Senators could be elected by AV.

[jimrtex]

MN, with nestings
quota for Senate is 85,172; for House, 42,586.

Clearwater+Mahnomen+Becker+Clay+Otter Tail: 2 Senate, 4 House

I would score the Senate plan on (1) number of small county cuts; (2) modest large county cuts; and (3) population equality. There could conceivably be a bonus if a house district can be divided at a county line (e.g. a 2 district senate cluster of smaller counties would require a county split for two house districts, but one or both of the other two districts might be made of whole counties.


Clearwater+Mahnomen+Becker+Clay+Otter Tail: This is one type (1) cut. I would want a demonstration that the cut (presumably Becker) is plausible in not requiring a township split. You might be able to move the cut to Clay or Otter Tail. It would then come down to communities of interest - but would not score differently so long as only one county was cut.

Your instincts are sharp. I found only one way you could avoid a township split in that cluster: have Detroit Lakes be in the same House District as Clearwater and Mahnomen counties.
https://davesredistricting.org/join/2827c85b-a434-472f-bf23-64ab3c01821e
Here's a DRA link for you to track my progress in mapping out how this would translate into an actual House map.

I cheated.

You might recall a lawsuit was filed (seemingly prematurely) in Minnesota last February alleging that the state was malapportioned, and that the legislature was unlikely to act.

https://talkelections.org/FORUM/index.php?topic=374399.msg7967230#msg7967230

Minnesota has a long history of failing to reapportion. The legislature has not adopted an apportionment plan for the legislature in its first session after the census since 1881 (sic). With split control it is reasonable that they will fail again this year, or even if they pass a map, the governor might veto it.

The purpose of the lawsuit was actually to establish jurisdiction with the Minnesota Supreme Court (after filing in a district court, they appealed to the Supreme Court, claiming that there could be numerous lawsuits in different district courts, and that therefore the Supreme Court should take over. Which they did - but saying we won't actually do something (unless/until) the legislature fails.

A special panel has been appointed:

Minnesota Judicial Branch Special Redistricting Panel 2021

There is a Minnesota statute that requires redistricting to be completed 25 weeks before the primary (mid-February), so the judicial panel has said they won't act until then. Meanwhile, they have to prepare for action. This gives an incentive for legislative parties to not compromise.

The initial complaint gave information about the plaintiffs (to establish their standing and interest). One had been involved in NCSL, and I was sure that he must know Muon, so I looked further at his submission. It had included some example plans based on population estimates, and also suggested standards that a panel should use. Among these was that the panel should use a +/- 2% limit (which the panel has now adopted).

I was going to produce a counter-example which showed that +/-5% could produce plans that respected county boundaries (and townships) and that there were objective standards for judging overall equality. I posted it beginning here:

https://talkelections.org/FORUM/index.php?topic=374399.msg7972615#msg7972615

I got side-tracked looking for precinct data, and other things, and never completed my effort. My map was based on my projections of Census estimates to 2020.

Anyhow, your map prompted me to revisit my map with the 2020 populations.



It appears that based on my projections I was able to put two whole districts in Saint Louis. Duluth was one senate district. The other was the Iron Range and northern portion of the county.

The partial district snaked through between the two connecting Carlton to Lake (and Cook). The population in Lake and Cook is extremely concentrated along the Lake Superior shoreline, so it sort of like a Miami-Dade district extending in a long skinny line along the Florida Keys.

https://talkelections.org/FORUM/index.php?topic=374399.msg8001756#msg8001756

I had also remembered trying to get the district east of Fargo (or Moorhead) right. If you tried to place Fergus Falls or Detroit Lakes together with Moorhead, it would be overpopulated, even if you played around with Wilkin.

This shows my split which produces a sort of stubby peninsula between Fergus Falls and Detroit Lakes, that is entirely in Otter Trail.

https://talkelections.org/FORUM/index.php?topic=374399.msg8001425#msg8001425

My map appears to be proto-clusters. So that while Clay-Wilkin is underpopulated and Otter Tail-Becker is overpopulated, the two complement each other.

Based on my projections, Stearns would be just above 1.900 and Olmsted would be just below 1.900, and the reverse is true, so I will also have to make an adjustment there.

Also the brown Brown district in the southwest is about 10% over. I intended to pair it with the far southwestern green district.

So I will update to form more conforming clusters.

[jimrtex]

Update: I now have a complete House and Senate map. These lines are far from necessarily final. It can be seen through means of the DRA link I posted on the previous page in the quote chain with Jimrtex.
Average deviation is 2.64%. Overall deviation is 9.94%.
In the House, there are 60 seats that went for Trump in 2020, and 69 that went for him in 2016.
In the Senate, there are 32 seats that went for Trump in 2020, and 36 that went for him in 2016.

These are my cluster maps.





Under the Texas Constitution, districts should be multi-county single-member, or single-county, one or more members. There are also floterial districts which are apportioned a single representative for zero or more whole smaller counties and the surplus from one or more larger counties - where the surplus is the excess above the number of whole quotas.

Floterial districts are appropriate for apportionment purposes - they violate equal protection when used for electoral purposes, since the entire electorate of a larger county may vote for the representative, even though the county only contributed its surplus to the apportionment process.

The Texas Constitution is harmonized with equal protection by taking an area containing the surplus population and attaching it to other surplus areas or whole counties.

Smaller counties may never be split under the Texas Constitution, it may be necessary to do so in order to comply with equal protection. Such cuts should be kept to the minimum possible. "possible" and "necessary" are used in a mathematical sense. Any plan with fewer small county cuts defeats any plan with more small county cuts so long as the overall plan is within a 10% range.

Division of a surplus into two or more areas is treated the same as a small county cut. A county can only have one surplus, which is always less than a quota.

Larger counties should be divided into as many whole districts as possible. These are not counted as cuts since they respect the county boundaries.

Detaching an area equal to the surplus of a large county is relatively benign so long as it produces better overall equality. In general, the surplus of the combined area should be less than the surplus(es) of the contributing larger counties.

Scoring of my plan:

One or more whole counties, single district, no penalty: Polk (listed by largest county in district), Beltrami, Itasca, Crow Wing, Douglas, Isanti, Chisago, Renville, Nobles, Brown, Blue Earth, Mower, Winona, and Goodhue. (14 districts)

One whole county, multiple districts, no penalty: Hennepin (15), Dakota (5), and Olmsted (2).

Optimal large county splits:
Ramsey(6)-Anoka(4)-Washington(3), one shared. 3 large county splits.
Scott(1)-Carver(1), one shared. 2 large county splits.
Saint Louis(2)- one shared with Carlton, Lake, and Cook. 1 large county split. (this presumes that Carlton and Lake may be connected by a corridor between Duluth and the Iron Range.

Small county splits:
Clay one cut, likely in Otter Tail (small county cut)
Rice one cut in Rice.

Large county cut surpluses:

Wright. One district in Wright (1 large county split), split of Wright surplus, with one extending into McLeod+Sibley, the other into Meeker+Kandiyohi

Stearns. One district in Stearns (1 large county split), One district in Sherburne (1 large county split), and one additional split, likely of Sherburne.

Total: 4 small county splits. 9 large county splits. The large county splits are mostly disregarded.

42 districts entirely within a single county, 14 single-member multi-county districts.
11 districts containing parts of one or more counties.

Standard deviation assuming perfect splits within clusters: 2.05%

Note: My projection had Olmsted below 1.900 and therefore required augmentation. It happens that the the overall deviation range is 9.88% even with the two adjacent districts at +5.5% and +5.4%.

My projection had Stearns above 1.900 and therefore could have had two districts. I had to combine it with the district across the Mississippi River, and requires an additional small cut.

Your plan splits 6 small counties, and the surpluses of Stearns and Washington. It also splits the surplus in Saint Louis, but that may not be necessary. The standard deviation for your plan is 2.40%.

If ours were the competing plans, my plan would win based on fewer small county splits: (4) vs. (8).

We could then turn the plan over to county districting commissions.

[jimrtex]

We could then turn the plan over to county districting commissions.

If you were drawing a House map using these same clusters, how would you go about it? I have the feeling the clusters I used are slightly House-optimized (designed to produce whole-county HDs in rural areas and less seats crossing county lines in more urban ones). SDs would be composed of pairings of whole-county HDs in greater Minnesota.

An example of this is the Pope+Stevens+Wilkin+Traverse+Big Stone+Lac qui Parle+Yellow Medicine+Lyon+Lincoln cluster, an unsuccessful example was the Pine+Kanabec+Aitkin+Mille Lacs cluster, which was supposed to neatly split into two whole-county HDs both taking two counties, but my brain screwed things up and I found myself having to work with a mistake. Of course, it wasn't a bad mistake, and whole-county pairings are not necessary easily to come by in the numbers.

I'm not a fan of legislative nesting. If you require a high level of population equality, some districts are going to be pretty good, and others quite ugly. Communities of interest do not come in quantum sizes. You are not necessarily going to end up with something better by splitting a district in half or joining two together.

In a state like Minnesota with a large senate, there is less differentiation between the two bodies. What is the point of having two duplicate bodies, where the A and B representatives are quite likely to be the same party as the that of the senator.

So first what I would do is draw the senate districts in the larger magnitude clusters taking care to maximize the number of senate districts wholly in Saint Louis, Ramsey, Anoka, Washington, Scott, Carver, Wright, Sherburne, and Stearns. I will pay attention to township/city boundaries, and to an extent to population equalization.

The drawing of house districts would be treated as 67 separate activities. I might back out some senate districts in larger counties if I can see a way to make better house districts with an adjusted senate configuration.

I should be able to draw a house district wholly in Blue Earth, Winona, Rice, Isanti, Clay, Otter Tail, Crow Wing, Chisago, and Kandiyohi. I might be able to do so in Itasca, Beltrami, and Goodhue.

Kandiyohi and Isanti should be whole county districts. While Becker has the correct population it won't be. If any multi-county house districts fall out, I'll accept them. I think it will be around 100 house districts that do no split a county, which is about 3/4 of the total.

[jimrtex]

If you were drawing a House map using these same clusters, how would you go about it? I have the feeling the clusters I used are slightly House-optimized (designed to produce whole-county HDs in rural areas and less seats crossing county lines in more urban ones). SDs would be composed of pairings of whole-county HDs in greater Minnesota.

An example of this is the Pope+Stevens+Wilkin+Traverse+Big Stone+Lac qui Parle+Yellow Medicine+Lyon+Lincoln cluster, an unsuccessful example was the Pine+Kanabec+Aitkin+Mille Lacs cluster, which was supposed to neatly split into two whole-county HDs both taking two counties, but my brain screwed things up and I found myself having to work with a mistake. Of course, it wasn't a bad mistake, and whole-county pairings are not necessary easily to come by in the numbers.

This might be an example.



In my previous effort I had split the four county area around Moorhead, as Clay+Wilkin and a peninsula running north of Fergus Falls in Otter Tail. I couldn't get to Detroit Lakes or Fergus Falls particularly if trying to balance the population. So instead I tried to avoid Detroit Lakes and Fergus Falls.

In a rural area like this I would ordinarily seek perfect balance without splitting townships.

But I noticed that Moorhead could form a largish House district. This meant that I had to make the Moorhead senate district a bit larger so that the second House district would still be above the lower threshold. At the same time, I had to make sure the Otter Tail senate district was not too small.

Moorhead senate seat: -0.2%
Moorhead house seat: +4.6%
Other House seat: -5.0%

Other senate seat: -3.6%
This will in turn force the two House seats to be pretty equal so as to make sure both are in bounds.

If I try to keep one House district entirely in Otter Tail, which I will, the other district will have to form a sort of question mark shape from Becker, through northern Otter Tail and down to Wilkin.

[jimrtex]

I completed the senate map except where I was going to have to split cities or townships. After I downloaded VTD's I decided to start splitting into House districts.





SD-1 was easy since the split can be done on county boundaries and HD-1A is within 16 persons of the ideal population (the A house district is the more saturated color).

I was going to try to create 2A and 3A in Beltrami and Itasca counties since they constitute a majority of the population in the putative senate districts. But since they were in the middle of the senate districts there was going to be a corridor connecting the northern and southern portions of the house districts. These corridors would cut through Indian Reservations.

If I were starting over, I would treat the reservations as county equivalents in defining clusters. But I didn't start over. I added the portions of White Earth in Becker County to my SD-2 cluster, but then decided to try to create a house district including White Earth, Red Lake, and Leech Lake. Which I have done.

I then drew HD-2B to include the Bemidji area, mostly in Beltrami, but extending into Hubbard and Clearwater. This fits pretty well with HD-2A to form SD-2.

This then required HD-3B to be drawn across the southern parts of Hubbard, Cass, and Itasca. HD-3A includes Grand Rapids and the western portion of the Iron Range.

Overall, these four districts are on the high end of population: HD-2A 1.043, HD-2B 1.034, HD-3A 1.031, and HD-3B 1.026.

[jimrtex]

This is Senate districts 4, 5, and 6, along with associated House districts. In particular I was demonstrating that two whole districts could be drawn in St. Louis, with the remnant attached to Carlton, Lake, and Cook counties.



It happens that Duluth is just the right size for a senate district. The house split is quite natural. Presumably the area to the southwest closer to harbor is more working class, and the area to the northwest out along the shoreline is leafier (though perhaps this is more coniferous).



The Carlton-St Louis-Lake-Cook senate district is a bit odd, but it should be remembered that the Lake-Cook population is concentrated on the Lake Superior shoreline. Rather than an odd bit of Saint Louis attached to the other counties, House 5B should be understood as Duluth suburbs extending out along the shoreline (similar to how the Florida Keys attach to southern Miami-Dade. Under 2/5 of the House district is in Lake and Cook.

House district 5A is based in Carlton. It extends northward into Saint Louis to include all of the Fond du Lac Band of Lake Superior Chippewa Reservation. A small area in northern Saint Louis is included in HD-3A to place all of the Bois Forte reservation in that House district. This extension was not shown on the map of 3A.

SD-4 in northern Saint Louis is divided into HD-4A which is based in the densest populated area of the Mesabi Range (Hibbing to Virginia-Evelyth). It can not be extended further east due to population reasons. A more global division of Saint Louis into northern and southern parts would divide the Mesabi Range. HD-4B can be comprehended as "Saint Louis excluding Duluth and much of the Mesabi Range".