North Carolina Legislative Redistricting

[jimrtex]

Bartlett v Strickland (2009) was a SCOTUS decision that affirmed a North Carolina Supreme Court decision that a minority group must constitute a majority in an area before Section 2 of the VRA required creation of a minority opportunity district.

In 1993, North Carolina had drawn a district (HD-98) as a majority BVAP district that was quite ugly, extending across Columbus, Brunswick, New Hanover, and Pender.



HD-98 is the rust red district. It includes three point-intersections to permit districts to cross and maintain contiguity (the North Carolina Supreme Court has since outlawed this technique). There are six other places where connectivity is less than a 1/4 mile (the district is about 63 miles tip to tip).

Two of the intersections permit HD-14, the pink district, to cross over to include some less black areas, and the other permit HD-98 and HD-96 the aqua district to cross over each other.

HD-12 is 27.3% BVAP
HD-96 is 24.2% BVAP
HD-98 is 55.7% BVAP
HD-14 is 12.5% BVAP


As proposed in 2001, HD-98 had been renamed to HD-18.



The boundaries are smoother, but the BVAP had dropped:

HD-19 25.7% BVAP
HD-18 40.3% BVAP
HD-17 14.7% BVAP

While smoothing had some effect, the 1990s district had dropped to 46.7% BVAP, and was 7% below the ideal district size. With the old district maximized to past 50% BVAP, the adjacent territory was likely to be much lower.

The North Carolina Supreme Court blocked this map in Stephenson I, which ruled that the map violated the whole county provisions of the North Carolina Constitution. The NC Supreme court drew an interim plan for the 2002 election that removed the intrusion into Pender, and increased the BVAP to 44.1%. The legislature proposed a new version in 2003, but that was also blocked by the NC Supreme Court in Stephenson II


The North Carolina finally accepted this map which was used for the 2004, 2006, and 2008 elections.



I'm having problems with projections but HD-17 and HD-20 are in Columbus and Brunswick; and HD-16, HD-18, and HD-19 are in New Hanover and Pender.

The old portion of HD-18 that extended westward along the northern boundary has been removed, and a larger area north into Pender has been added. The southern part of HD-18 is Wilmington in New Hanover County. About 2/3 of the population of HD-18 is in New Hanover, so we can see that the district was really the black areas of Wilmington, plus some rural areas to get the total population.

The districts are:

HD-16 7.4% BVAP
HD-17 11.0% BVAP
HD-18 39.1% BVAP
HD-19 5.4% BVAP
HD-20 28.1% BVAP

The coastal districts have a quite low BVAP% as these are just north of Myrtle Beach, SC and attractive to retirees from both the North and South who might find beachfront property in Florida too expensive.

New Hanover had 2.390 quotas, and Pender 0.612. Pender is almost evenly split between HD-16 and HD-18. So Pender which was entitled to about 3/5 of a representative, was chopped in half, and both fragments placed in districts with New Hanover, where New Hanover had about 2/3 of the population. Instead of a small majority of one district, it was a small fraction of two, dominated by its larger neighbor.

In 2004, the Pender County Board of Commissioners sued. Pender County later withdrew, but the suit continued with individual plaintiffs including former county commissioner Dwight Strickland with Gary Bartlett, the executive director of the North Carolina Board of Elections as the named defendant.

The plaintiffs argued that since HD-18 did not have a BVAP of 50%, that the district was not required by Section 2 of the VRA. Pender County was not covered by Section 5 of the VRA (around 40 counties in North Carolina were). Since the district was not required by federal law, it violated the whole county provisions of the North Carolina Constitution (as harmonized by the NC Supreme Court with OMOV provisions of federal law), which require minimum traversals of county boundaries.

The state argued that the district was required by Section 2, and was an effective cross-over district with a 39.9% BVAP. In 2007, the North Carolina Supreme Court ruled that since the district did not satisfy the Gingles Test it was not necessary to be drawn to satisfy the VRA, and thus it only needed to be drawn to satisfy OMOV. The SCOTUS heard the appeal in 2008, and upheld the NC court's decision in 2009.


The legislature redrew HD-16 and HD-18 for the 2010 election.



Red: Area in HD-18 in 2008 and 2010 (in Wilmington)
Orange: Area in Pender (and tiny part of Wilmington) transferred from HD-18 to HD-16.
Lime: Area in HD-16 in 2008 and 2010 (New Hanover-Wilmington and Pender)
Olive: Area in Wilmington transferred from HD-16 to HD-18.
Blue: HD-19 (unchanged).

The BVAP changes from 2008 to 2010:

HD-18 39.1% to 28.3%
HD-16 7.4% to 16.5%

The 2011 redistricting changed the county groupings, so that New Hanover is paired with Brunswick and Pender is paired with Onslow. HD-18 is now Wilmington plus an area in Brunswick. It is now 28.5% BVAP.


HD-18 elected Thomas Wright, a black Democrat, from 1992 through 2006. In 2008, he was expelled from the House in 2008, the first member to be expelled since 1880, for charges related to not reporting campaign funds and converting to personal use (this was in 6 figures). He was also convicted in criminal court, and served in prison from 2008 to 2014. He had refused to step down from his 2008 election effort, but his felony conviction knocked him off the ballot.

In North Carolina, vacancies are filled by the party of the former representative. The Democratic party chose a black Democrat who was then elected in November 2008. After the district was changed for the 2010 election, she decided not to run.

A white Democrat was elected by a 57%-43% margin in 2010, after she narrowly defeated a black Democrat in the primary. After the district was redrawn she was re-elected with 67% (2012), unopposed (2014), and 61% (2016). She resigned this year to take a cabinet position, and was replaced by another white Democrat.

[jimrtex]

Earlier I had produced this map, dividing the state into whole county groups, each entitled to a whole number of representatives.



Under the state constitution, counties may not be divided in creation of representative districts. The constitution does not forbid multi-member districts, so the above would be a constitutional map, that overall minimizes the size of multi-member districts.

But the US constitution as interpreted by the SCOTUS frowns on the use of multi-member districts where there are racial or possibly political minorities involved. If Wake was a 11-member district and Mecklenburg a 12-member district, there could be wild partisan shifts statewide if the Democrats or Republicans carried the two counties, and black Democrats might be shut out in the primary.

Thornburg v Gingles (1985) was based on challenges to multi-member districts for the North Carolina House: Mecklenburg (8 representatives); Forsyth (5); Durham (3); Wake (6); and Edgecombe, Nash, Wilson (4); and Senate: Mecklenburg, Cabarrus (4). The SCOTUS reversed the decision with respect to the the Durham House district.

The SCOTUS devised the Gingles test for multi-member districts:

(1) The racial or language minority group "is sufficiently numerous and compact to form a majority in a single-member district";
(2) The minority group is "politically cohesive" (meaning its members tend to vote similarly); and
(3) The "majority votes sufficiently as a bloc to enable it ... usually to defeat the minority's preferred candidate."

This might reasonably still be applied to the large multi-member counties, since by the North Carolina they should be multi-member districts. Cities are also likely to be more racially segregated on a large scale. In more rural parts of the state it is more challenging. What are the proper areas to consider whether a majority-minority district can be formed, is a district strung out across three or more counties compact, particularly if single-member districts can be formed in areas of one or two counties?

If BVAP is over 40%, then (3) will not be true, since if 1/6 (or less) of the white voters will vote for a black Democrat, they are no long voting sufficiently as a bloc to defeat the minority's preferred candidate. But if you can only get to 40% BVAP in a reasonably compact area, the first condition is not true.

After Gingles, North Carolina did switch to smaller districts, but the North Carolina Supreme Court has determined in Stephenson that even two-member districts should not be used unless there is a compelling interest in doing so. Their finding was based on a belief that in a two-member district, local interests might be submerged, or that voters from two-member districts would have more influence in the legislature.

The use of single-member districts and One Man, One Vote requirements (maximum of 5% deviation requires splitting of some counties.

North Carolina has 9 counties that have multiple representatives, where the districts can be drawn entirely within the county: Mecklenburg (12), Wake (11), Guilford (6), Cumberland (4), Buncombe (3), Davidson (2), Iredell (2), Catawba (2), and Alamaance (2), for a total of 44 of the 120 House districts (36.7%).

North Carolina has 3 counties that can have a single district coincident with the county: Caldwell, Wilson and Lincoln. These could also be considered as part of group of single counties that can have one or more whole districts; or a group of one or more counties that can have one whole district.

North Carolina has 23 counties entitled to more than one district, but can not have all districts contained in the county. Collectively the counties are entitled to 43.93 districts, but only 35 whole districts. The average surplus for these 23 counties is (43.93 - 35)/23 = 0.388. That this is significantly less than 0.500, is due to many of the counties being entitled to one district plus a fraction, with a bias toward the lower end. These counties must be divided, with a fragment attached to adjacent counties in order to comply with One Man, One Vote and the North Carolina constitution. The rest of the county should be formed into as many whole districts as will fit into the county so as to avoid county line traversals.

North Carolina has 65 counties entitled to less than one district. They must be combined with other counties. If the counties in a county group are all small, and the county group is entitled to one district, then that district is entirely legal under OMOV and the North Carolina constitution, regardless of the number of counties. The 65 small counties have a population equivalent to 29.04 districts.

The surplus of the 23 large counties (9 districts) and the 29 districts equivalent to the 65 small counties represent the maximum number of county groups that could be be formed from these counties. Add in the twelve single-county groups, the maximum possible number of groups is 50. My map has 43.

The difference between the maximum and the number of groups in a map represent counties that are divided for reasons other than the population of the county alone.

In my map the seven counties (actually excessive divisions) are:

Haywood (0.743 quotas) divided between two districts. This splits the county almost in half. The split looks ugly on a map, but keeps Waynesville, the only significant town, whole. The other part of the county does maintain a small majority of its district. Neither representative is from Haywood. While the split might not be good for Haywood, it may provide an opportunity for two very small counties, Swain (0.176 quotas) and Yancey (0.224) to elect a representative.

Wilkes (0.873 quotas) is divided between three districts. It is difficult to create groups in this area since Caldwell, Catawba, and Iredell, box Alexander in, and even though Alleghany and Wilkes could form a single member district they can't because of Alexander.

The group of Alleghany, Alexander, Wilkes, Surry, and Yadkin can almost form three groups (1) Wilkes and Alleghany qualify as a group, but can't be used because Wilkes cuts it off; (2) Surry almost has enough for a single-county group, but is just out of population range at -7.3% of a quota; and (3) Alexander and Yadkins could form a group, but are not contiguous. The districts are drawn to match the three almost-groups. A small portion (2.4%) of Wilkes is used to connect Alexander and Yadkins; and a small piece (4.2%) of Wilkes is added to Surry; the remainder (93.4%) of Wilkes is combined with Alexander.

Montgomery (0.350 quotas). Most of the county is paired with Stanly, and the remainder with Randolph. Montgomery is currently divided, though a different division is used.

Sampson (0.798 quotas). Sampson, Bladen, and Duplin form a group with two representatives, forcing a split of Sampson. Sampson will have a small majority of the district formed with Bladen.

Robeson (1.688 quotas). One district will be in Robeson, with the rest of the county split between one district with Hoke, and one district with Columbus. Instead of being a dominant part of a second district, Robeson will be a minority in both. The district in Robeson can form a Native American (Lumbee) VRA district.

Harnett (1.443 quotas). One district will be in Harnett, with the remainder split between one district with Lee, and another with parts of Wayne and Johnston.

Rockingham (1.178 quotas) is part of a group with Stokes to the west and Caswell to the east. Currently Rockingham is divided along an odd boundary, that placed Reidsville with Stokes, but not the residence of the representative of the other district in Reidsville. My understanding is that ideally, one district should be entirely in Rockingham, with Stokes and Caswell connected by a narrow corridor. This corridor would have to be along the southern boundary, since the largest city, Eden, is right on the northern (Virginia) border. It appears possible to snake a corridor along the southern edge of Rockingham, though it would involve splitting all four precincts along the line.

[jimrtex]

[...]
The difference between the maximum and the number of groups in a map represent counties that are divided for reasons other than the population of the county alone.

In my map the seven counties (actually excessive divisions) are:
[...]

Interesting that the number of counties with excessive deviations is equal to the difference between the maximum "ideal" number of whole-county, whole-district conglomerates and that in your map.  The Wilkes conglomerate causes a 2-group reduction, while the fact that Rockingham County has two partial districts rather than one whole and one partial district doesn't cause any reduction in the number of groups.

I think (correct me if I'm wrong) that your calculation of the maximum (or "ideal") number of county groups is equivalent to the following calculation with a very different heuristic: take the size of the body in question, then subtract the number of counties that are too big for one district (so, in the US, those counties with a quota over 1.05), then subtract the number of counties that are too big for two districts (quota over 2.1, and that includes the counties for which you've already subtracted one for), then subtract the number of counties that are too big for three districts (quota over 3.15), and so on.  Or you could also think of it as taking the size of the body, subtracting 1 for counties with quotas between 1.05 and 2.1, subtracting 2 for counties with quotas between 2.1 and 3.15, etc.  My heuristic may be more complicated, but basically counties that are in the "1+1" (Trond-speak) population range have the same effect on the maximum or "ideal" number of county groups) as those that can have an even 2 districts, and 2+1 the same as 3, etc.  And it doesn't matter if you combine two large counties in the same conglomerate, as the loss of one of the two "remainder districts" is made up for by (ideally) another single district consisting of whole counties.

Androscoggin County, Maine increased the size of its Board of Commissioners from three to seven effective for the 2014 primary and general elections (the new commissioners took office in January 2015), in its charter approved by Androscoggin County voters in November 2012.  The quotas for each town in the county (as of and according to the 2010 census) were as follows:

Auburn city 1.4984
Durham town 0.2501
Greene town 0.2827
Leeds town 0.1512
Lewiston city 2.3783
Lisbon town 0.5855
Livermore Falls town 0.2071
Livermore town 0.1362
Mechanic Falls town 0.1970
Minot town 0.1694
Poland town 0.3494
Sabattus town 0.3169
Turner town 0.3727
Wales town 0.1050

The maximum number of groups consisting of whole towns and whole commissioner districts (and not consisting of any smaller such groups; you could call them "minimal whole town groups" or "minimal whole county groups" as applicable) is 7 - 1 (Auburn) - 2 (Lewiston).  Auburn and Lewiston are across the Androscoggin River from each other (the two cities merging has been an on-again, off-again topic for decades, but I've heard there might be a binding referendum on a merger this fall, or maybe it's next fall) and have remainders that add up to less than 1.  So, "ideally", you could have either a 3-district conglomerate including Lewiston, a 2-district conglomerate including Auburn, and two single-district conglomerates of whole towns, OR a 4-district conglomerate including both Lewiston and Auburn and three single-district conglomerates of whole towns.  It turns out only the first was possible.  Lewiston and Auburn would have been within range of four districts either just by themselves (-3.08%) of with Durham (+3.17%), and the reminder of the county would have been within range of three districts in either case (+4.11% or -4.23%), but you wouldn't be able to break down the "rural" 3-district conglomerate even into a contiguous 1-district and 2-district conglomerate (where single districts of whole towns could be formed that were within range, either areas would be cut off of the remaining two-district conglomerate would be over- or underpopulated).  What ended up being done was Lewiston formed a 3-district conglomerate with Durham and Greene (on opposite sides of Lewiston, and Durham only borders Lewiston across a fairly small section of the Androscoggin River with no bridge anywhere near, connecting Durham to Green via eastern Lewiston looked fairly easy though) and Auburn forming a two-district conglomerate with Poland and Mechanic Falls.  Lisbon-Sabattus-Wales (the neatest pairing by far) and Livermore-Livermore Falls-Leeds-Turner-Minot districts rounded out the plan.

This rule applies to West Texas in determining the number of rural districts (without an anchor of Potter, Randall, Lubbock, Ector, Midland, Tom Green, Taylor, or Wichita). Lubbock+ must form two districts, and each of the other 7 must form the core of another district. Total population of the area (in quotas) minus (9 = 2 + 7)

In Texas, HD-88, HD-68, HD-60, HD-59, and HD-53 form five groups, which have to exclude HD-69, HD-71, HD-72, HD-81-84, HD-86 and HD-87. HD-68 and HD-88 don't have to be that extremely drawn, but there was a reduction of one such district, which forced a pairing of two Republicans. One from the Panhandle agreed not to run for re-election (he later ran for the Railroad Commission). This permitted his district to be divided between HD-68 and HD-88. The representative for HD-68 is from Cooke County.

How many total depends on what happens with any counties taken up in districts anchored by counties on the I-35 corridor (Johnson, McLennan, Bell, Williamson, Travis, Hays, Comal). Ideally you should also divide the regions so that the population of a region is proportional to a whole number of districts. In Texas, you can decide which way to go from I-35 to better balance east and west Texas. This also plays into determining the boundaries between south/border Texas and east Texas.

I don't know whether you have ever painted yourself into a corner by starting out drawing districts from Aroostook or York, and ending up with a population equivalent to 2.3 districts and two more districts to draw. This could easily happen if the average population of the other 149 House districts was (151 - 2.3)/149 or 0.998 of a quota. This is one reason that I always normalize population. You are less likely to have a consistent bias if your target is 1.000 than if it is 7386. Your mind (or my mind at least) will think that 7050 is closer to the quota than 7722 is. 7386 is somewhat over 7000, and 7050 is a bit over 7000, bu 7722 is close to 8000.

Where I am going with North Carolina is that the state expert claims that the optimum map has 41 county groups, rather than the 43 that I achieved. In addition, his plan change areas of the state that are not being litigated over, and thus will result in unnecessary redrawing of districts. What is even more interesting is that I adopted the current plan in those areas. I started out with the counties that could form a single-county group and then started drawing west from Charlotte. Because of the restricted area there weren't many alternatives. I then found the current map, and figured out the five-county group centered on Haywood. But I don't think that I would have discovered the 5-county Wilkes group.

As I did the rest of the state, I tried to maintain regional parity, and may have found a better plan because of that. The "optimum" plan from the state's expert, has a seven-county group Lee-to-Greene and tail to Sampson and Duplin with a population of 7.349 quotas (or 1.04986 quotas per district). This will force redistricting down to the block level.

A plaintiff's expert has a computer program that randomly creates plans from possible county groups (there are a lot of groups with few counties, the challenge is getting them to fit together). His best effort, out of 2000 tries was also 41. He was trying to shoot fish in a barrel. The current plan has a 20-county group from Stanly to Dare because of the linking of VRA groups.

There North Supreme Court appears to have messed up. The constitution says that "No county shall be divided in the formation of a representative district". The correct harmonization should be:

The number of counties that are divided for reasons other than to provide whole districts within a county should be minimized.

But it appears that it has instead stated an algorithm: Form single-county groups; form maximum number of two-county groups; ...

While two-county groups ensure that they have no excessive cuts, it ignore that small counties that form a group of any size are entirely constitutional.

My approach to Androscoggin might be to use a population of 0.3783 and 0.4984 for the two cities. This could lead to better population balance since you will be limiting the totals to a range of 0.95 to 1.05. Or maybe not. It is similar to the Texas rule that surpluses must be placed in a single district, and where you have agglomerations of such counties (DFW), Houston, I-35 corridor, there might be limited options.

The hard part will be the name for the merged city. When Mindspring and Earthlink, there were suggestions for the merged company, such as MindLink, EarthSpring or Mindlink Earthspring. I preferred the anagram: Leningrad Knit Shrimp

Some possibilities: Unwise Urban Lot, Blue Straw Union, and Raw Subtle Union, Uninsurable Two, and Worn Unsuitable

[jimrtex]

Associated with the North Carolina litigation is an analysis of the whole county groups.

Supplementary (plaintiffs) material for Dickson v Rucho(PDF)

Dickson v Rucho is the congressional and legislative redistricting case tried in North Carolina courts. The North Carolina Supreme Court upheld both plans. The same lawyers then recruited a different set of clients to challenge the plans in federal court.

The congressional redistricting federal case is Cooper v. Harris which the SCOTUS affirmed earlier this year. The North Carolina legislature had already drawn new congressional boundaries which were used in 2016. Those boundaries are still being litigated.

The legislative redistricting federal case is North Carolina v. Covington. The SCOTUS affirmed this decision after the congressional district case. While it was briefed, the SCOTUS did not hold oral arguments, suggesting that they believe the same principles were at play in both cases. However, the SCOTUS vacated the order that would have required a truncation of legislative terms, and holding special elections in 2017, implying that district court had only examined the issues in a cursory fashion. This is being litigated now.

The SCOTUS has also overturned Dickson v Rucho, and sent it back to the North Supreme Court so it matches the logic of the federal cases. I don't know how much has to be done, or whether it is effectively moot at this time. One problem is that the North Supreme Court has directed that VRA districts be drawn first, before drawing whole county groups, and the federal court was evidence of race predominating over other factors, such as in this case adherence to political subdivisions.

Personally, I think if you have to ignore things such as political subdivisions, the districts are inherently not compact.

Back to the PDF file. On Page 26 (PDF) is the affidavit of David Peterson, an expert witness for the plaintiffs who has written a computer program to find optimum whole county groupings. He suggests that rather than trying to find the largest number of small groups, one should concentrate on trying to avoid groups with large numbers of representatives.

Both are wrong. One should try to create the maximum number of groups since that will minimize the number of cuts needed to equalize district populations. Any other cuts are necessary in order to create single-member districts. Part of Peterson's role is to demonstrate how inefficient the cover used by the legislature is. But that is not hard to do since they ended up with a 20 county group stretching from Stanly to Dare to fit all the VRA snakes into the group.

What is being confused is that both groups with fewer representatives and fewer counties contribute to having more groups.

There are 120 representatives. If the average number of representatives per group is smaller, it means that there are more groups. My plan has 120/43 = 2.791 representatives/group. To beat it you need to create more 1 and 2 member groups.

There are 100 counties. If the average number of counties per group is smaller, it means there are more groups. My plan has 100/43 = 2.326 counties/group. To beat it you need to create more 2 county groups (since I already have the maximum possible one county groups).

On PDF Page 32 there is a chart showing the best plans found by Peterson's program. Three of them are superlative in one characteristic or another.

Plan 1211 has 41 groups. My plan has 43. Moreover, I have more 1-representative and 2-representative groups AND more 2-county and 3-county groups.

Plan 1571 has 15 one-representative groups, one more than mine. But it has three fewer groups, four fewer 2-representative groups. It also has four groups with six or more counties. My largest groups are five counties, none of Peterson's achieve that.

Plan 1579 has 17 two-county groups, one more than mine. But it has three fewer one-representative groups. It also has four groups with nine or more counties, including a monster with nine.

       |  Representatives  |      Counties
Plan GR |  1  2 3 4 5 6 7 8 |  1  2 3 4 5 6 7 8 9
JIMR 43 | 14 12 6 5 2 2 0 0 | 12 16 7 5 3 0 0 0 0
1211 41 | 13 11 6 4 3 1 1 0 | 15 15 6 4 1 2 1 0 0
1571 40 | 15  8 5 3 5 1 0 1 | 15 15 5 4 2 1 1 0 0
1579 40 | 11 13 5 5 1 2 0 1 | 12 17 4 2 1 2 1 0 1

Beginning on PDF Page 52 there are details of each of Peterson's plans (or covers). Plan 1211 is on PDF Page 85 and 86.



If you skip down in the Super District ID column, you will come to Super District 1, consisting of the single county of Alamance with 2 representatives. Peterson's program built a library of possible super districts (or county groups) that have a population equivalent to whole number of representatives (+/- 5%).

Super Districts 1 to 12  are the 12 single-county groups in the state. These are practically mandatory. There could be conceivably such groups place in a way that other super districts can not be formed, in which case you would have to place such a county in a multi-county groups. It is conceivable that such a placement could improve overall performance of a plan, but I very much believe it would be accepted.

Super Districts 13 to 55 are the 43 two-county groups, in alphabetical order from 13 Alleghany-Wilkes to 55 Richmond-Scotland.

Peterson does not utilize 13 Alleghany-Wilkes; 31 Cleveland-Rutherford; and 33 Davie-Forsyth. 13 and 31 cannot be used since that would isolate Alexander and Gaston, respectively. I don't know why they don't use SD 33. I use it, and it is used in the current map.

Three-county groups begin with 56 Alexander-Watauga-Wilkes. Many are not used. I know that 114 is a three-county group (Johnston-Nash-Wayne) and 129 is a 4-county group of Anson-Montgomery-Randolph-Richmond.

Potential super districts that contain a smaller super district are apparently excluded. For example since Beaufort-Craven form a two-county super-district, then Beaufort-Craven-Pamlico and Beaufort-Craven-Pitt are excluded. I think this is a bug. While if an entire super district could be divided into smaller super districts (example Burke-Cleveland-Gaston-Rutherford into Burke-Rutherford and Cleveland-Gaston), it makes sense not to use the larger group, this should not be the case if only part of the larger group can be treated that way. This might be why my plan is better than Petersons.

Finally, go back to the super districts with negative numbers. These are apparently when the program got stuck, and could not reduce an area further. The numbers are assigned on ad hoc basis, with the same set of counties, being given a different number for different covers.

In the map, these usually large groups, are depicted in red or orange. The group along the Tennesssee line of Haywood-Jackson-Madison-Swain-Yancey is used in my plan and almost all other plans. The single-county groups are in green, and the groups that are also used my plan are in blue. The groups in yellow are groups that are in plan 1211, but not my plan.

The standard deviation for Plan 1211 is 3.24%, assuming perfect splits of all superdistricts vs. 2.82% for my plan.

It is trivial to find where the extra cuts are made, by adding the surplus for the larger counties or the whole population for the smaller counties and rounding to the nearest integer. If this is 0 or 1, there is no extra cut needed.

[jimrtex]

Sigh! My plan provides for 121 districts.

The multi-county groups had population for 72.970 districts (just short of 73) and I created groups entitled to 74 districts (an error of -1.4%)

[jimrtex]

Sigh! My plan provides for 121 districts.

The multi-county groups had population for 72.970 districts (just short of 73) and I created groups entitled to 74 districts (an error of -1.4%)

Did you travel to Vermont recently?  (Do you remember what I'm alluding to here?)

Which of the following from your post where you talked about the number of groupings is incorrect?  The 43.93 districts for the "large" counties that aren't within range of an integer number of districts or the 29.04 districts for the "small" counties?  Because those two numbers add up to 73.97, not 72.97.

Was a good portion of the error from the districts and groupings you left as is (because the districts or groupings weren't overturned - not that a court would overturn or uphold a grouping per se but where the simplest solution seemed to be to keep the groupings)?

43.93 + 29.04 is 72.97. The groups should have had 73 districts. But because they were slightly small on average (0.986 quotas) I ended up with a total of 74 districts. My spreadsheet has sums for the total populations (120.00 quotas) and counties (100), but where I should have totaled the whole number of representatives (and checked for 120), I had simply rounded 120.00 to 120.

I thought I had been careful. I had checked the number of districts for the single-county districts before I even began. It is quite close to a whole number (47.03). Wake and Mecklenburg both have fairly large errors, but they cancel (11.573 + 11.339 = 22.912, and 12 + 11 = 23). So while Mecklenburg is entitled to an extra 1/4 of a representatives, they get a whole 1.

As I worked my way east, I thought I was checking the total number of whole representatives vs. the total number of quotas, but since instead I was simply rounding the number of quotas they always matched.

When I was reviewing Peterson's affidavit, I noticed that when he checked potential covers, generated by random placement of superdistricts, he rejected covers that totaled 119 and 121. I thought, "Hah! If you crafted a map by hand you wouldn't have such problems!"

When I started analyzing Peterson's cover with the most superdistricts, the one with 41, I found it had 9 excess cuts, just as predicted by theory. But then I noticed I had eight excess cuts, despite having 42 districts. It took me a long time to find my mistake.

Where I had most of the problem is with districts with more than one representative. When you can get an area between 0.95 and 1.05 representatives, you are grateful to find something that is in range. But when the target is 2.85 to 3.15 there are lot more possible combinations that work. So it was generally super districts with more representatives that were the problem. I'll show this on another map.

I have a corrected map, but it is down to 41 superdistricts. It might beat other covers on some secondary tests.

[jimrtex]

This shows the error for maps in my original plan.



The percentages are not the relative error for the superdistrict, but the deviation from the ideal total for the superdistrict.

For example for the orange district (Union-Anson) east of Charlotte, has a population equivalent to 2.871 quotas, which is 0.129 quotas below the ideal of 3.000 quotas (0.129 quotas equals 12.9% of a quota).

If you were to total all of the deviations on the map (one per superdistrict, not one per county - it is easier to label every county, rather than one county per superdistrict) they will add to -100.0% or enough to create one extra seat.

Notice there is not too much error in the single-member superdistricts. There is also a cluster of four underpopulated superdistricts south of Greensboro and Raleigh, eastward from Charlotte, and another two along the coast near Wilmington.

Collectively, there is a group of 10 superdistricts with a population equivalent to 27.302 quotas that has 28 districts for a relative error of -2.5%. If we can rearrange these superdistricts so that 27 districts are apportioned, then they will have an average error of 1.1% and the correct total of 120 districts. An important aspect of this is that it preserves the cluster of single-member districts in the northeast, which is one of the key aspects of my map.

This is my revised map, showing the deviation.



It has 120 districts, but unfortunately only 41 superdistricts, as 10 superdistricts in the southeastern part of the state have been rearranged into 8 superdistricts. The other 33 are unchanged from my original map. It preserves six single-member multicounty districts in the east. It rearranges one from Richmond-Scotland to Hoke-Scotland, which is probably preferable from a racial standpoint, since Hoke has a larger minority population than Richmond. The three single-member multi-county districts along the Tennessee line are also kept.

Along with the three single-county single-member districts, these are the only districts that fully comply with the North Carolina Constitution and the the North Carolina Supreme Court's expressed concern about cutting smaller counties. A small county that can not constitute a house district is at somewhat of a disadvantage, and this is exacerbated by being minced with parts attached to other districts. When combined with other small counties, they likely share a common regional interest such as being rural, and there is less risk on one county dominating or being dominated.

The attachment of Jones to Duplin is optional. Jones could be attached to Craven as well, but this division provides better population balance, and it also moves a division of Pender further north, closer to the Duplin County line, keeping a larger share of Pender within a single district.

There is a North Carolina state highway connecting the two counties directly. Jones is somewhat separated from Onslow by a large pososin along the county line.

This is my revised plan in a more conventional form.



This is my analysis of my revised plan:



Green: County groups that are unchanged and that do not contain overturned districts. The districts do not need to be redrawn, and would not be contested if the court orders a special election. The groups contain 42 districts.

Generally, they are in areas with smaller black populations, though there are exceptions.

HD-71 and HD-72 in Forsyth (Winston-Salem) are both 45% BVAP. It was determined that it was impossible to reach 50% BVAP for both of them. One factor that the court used in determining racial predominance was use of a bright line 50%+1 test for its districts. Neither district was challenged by the plaintiffs (A VRA Section 2 challenge requires a district-by-district analysis, though it can be in the context of an overall analysis).

HD-23 (Edgecombe-Martin) and HD-27 (Halifax-Northampton) have 52% and 54% BVAP respectively. But because they are comprised of whole counties, they adhere to traditional neutral redistricting principles of respect for political subdivisions and the whole county provisions of the North Carolina Constitution. Neither district was challenged by the plaintiffs. While the court did not find them constitutional, no litigant suggested that they were unconstitutional.

Blue: New groups, but that don't contain any overturned districts. The districts will have to be redrawn, and in some cases, there may be racial concerns.

Union-Anson is currently a county group with three districts, two suburban districts that are on the Mecklenburg line, and another district that includes the remainder of Union plus all of Anson. Stanly and Montgomery a slightly too large (1.112 quotas) for a single district, but there is a district comprised of Stanly and most of Montgomery.

The four counties were combined in a single group to better balance population. Shifting a bit of the excess population from Stanly to Union-Anson will keep the basis outlines of the existing districts, with tweaking of boundaries to equalize populations. Three of the districts have BVAP of 12%, 12%, and 13%. The Union-Anson district has a BVAP of 24% due to Anson having a 48% BVAP. But combining Anson with Richmond with only reaches 36.3% and not quite enough for a district. Bartlett v Strickland does not require drawing a VRA district unless an area has BVAP%. The crux of the  current litigation appears that you identify the compact area simultaneously with determining whether there it is 50% BVAP, rather seeing if you can construct a connected area that is 50%+ BVAP that can somehow be characterized as being compact.

The Currituck-Dare-Hyde-Pamlico groups is only 11% BVAP. Dare (3%) and Currituck (6%) have particularly low BVAP population.

The Brunswick-Columbia group includes some of the area that contained a once 50% BVAP district that later was at the heart of Bartlett v Strickland The district had a focus in the city of Wilmington and then meandered across rural areas Pender, Brunswick, and Columbus which featured three separate point connections where districts crossed each other. Demographic change (white retirements toward the coast, and statewide growth requiring larger districts) and the elimination of point connections reduced the BVAP below 50%. Stephenson forced the district into New Hanover and Pender, and Strickland resulted in the non-division of Pender.

The current version of the district remains centered in Wilmington, but extends into Brunswick. The district is only 29% BVAP, and the Brunswick portion of the district is only 19% BVAP (overall, the county is 11%  BVAP). Columbus is 30% BVAP, and it should be possible to draw the district that includes Columbus and extends into Brunswick with a BVAP in the mid-to-high 20%'s. The Brunswick district should be quite similar to the existing HD-17, but will requirea small augmentation of its population.

The Cartaret-Onslow group replaces the Onslow-Pender groups. Two groups will remain in Onslow, with a portion along the eastern line being added to Cartaret. Onslow has a 16% BVAP. Replacing the 18% BVAP Pender with the 6% BVAP of Cartaret, reduces the BVAP of the group from 16.4% to 13.1%.

It should be able to maintain HD-15 pretty much as it is. HD-14 could be shifted to take up the area vacated by HD-16, and give up area to HD-13 coming in from Cartaret. There could also be some shifts. The incumbents for HD-14 and HD-15 are both near the edges of their districts. There is no risk of them being moved out of their districts, but it would be easy to move the boundaries away from their homes, and make the two districts more compact.

[jimrtex]

I see your new plan no longer keeps all 53 districts in groups without any overturned districts the same.  I can see why you rearranged the southern coastal area and the area just east of Charlotte, though, as getting rid of the shortages there was doubtless helpful in getting rid of that extra district.  I know earlier you had critiqued others for not keeping the "unchallenged area" as it was.

What had set me off on this quest was the claim made the legislature was going to have redraw (IIRC) 81 house districts, even though only 19 were found unconstitutional.With 53 retained groups, I thought that  would be reduced to 67 that would be redrawn, and some not materially redrawn, if at all.

The areas that I lost in my revised plan were on the periphery. Is still keep the block of districts in the western part of the state. But I am now down to 42 retained districts, requiring 78 (65%) to potentially be redrawn.

Incidentally, the North Carolina House and Senate redistricting committees will hold a joint meeting next Wednesday, July 26.

[jimrtex]

Back to the Peterson cover with the most groups (Cover 1211)



Green Groups unchanged from current, and with no overturned districts. 16 groups, 31 districts.

Blue Groups changed from current, with no overturned districts. 7 groups, 22 districts. Note the Halifax-Martin group splits up two groups which had unchallenged majority minority districts.

Yellow Groups unchanged from current which contain no overturned districts. 4 groups, 33 districts.

White Groups changed from current which contain overturned districts.  18 groups, 34 districts.

Overall up to 89 groups (74.2%) may need to be redrawn.

Standard deviation assuming perfect splits of whole county groups 3.24%.

Forced county cuts (9 total):

(1) Columbus. One fragment with Pender, one with Robeson.

(2) Rowan. One fragment with Stanly, one with Davie, one whole district in Rowan.

(3) Montgomery. One fragment with Anson-Richmond, one with Randolph.

(4) Haywood. One fragment with Jackson-Swain, one with Madison-Yancey.

(5) Granville. One fragment with Durham-Person, one with Vance.

(6) Sampson. One fragment with Bladen, one with Harnett-Johnston-Wayne.

(7) Harnett. One fragment with Lee, one with Johnston-Sampson-Wayne, oine whole district in Harnett.

(8) Wilkes. One fragment with Alexander, one with Alleghany-Surry.

(9) Surry. One fragment with Stokes-Rockingham. one with Alleghany-Wilkes.

Seven of the extra cuts are in smaller counties (entitled to less than one representative). Splitting such counties is arguably worse since it will be difficult to be elected from the fragments since many fellow county residents will not be able to vote for a candidate, and the fragments will have less influence over the election of "their" representative.

In addition, Wayne will not have one whole district and a fragment, but will have two fragments.

Number of groups per number of counties (ie there are 15 2-county groups)
1: 12
2: 15
3: 6
4: 4
5: 1
6: 2
7: 1

Number of groups per number of districts (ie there are 11 2-district groups)
1: 13
2: 11
3: 6
4: 4
5: 3
6: 1
7: 1
11: 1
12: 1

30 counties are in 10 single-member, multi-county groups. All such groups/districts are consistent with the North Carolina constitution. There are an additional 3 single-member, single-county groups.


For comparison my revised plan:



Green Groups unchanged from current, and with no overturned districts. 20 groups, 42 districts.

Blue Groups changed from current, with no overturned districts. 4 groups, 10 districts.

Yellow Groups unchanged from current which contain no overturned districts. 7 groups, 42 districts.

White Groups changed from current which contain overturned districts.  10 groups, 26 districts.

Overall up to 78 groups (65.0%) may need to be redrawn.

Standard deviation assuming perfect splits of whole county groups 2.62%.

Forced county cuts (9 total):

(1) Haywood. One fragment with Jackson-Swain, one with Madison-Yancey.

(2,3) Wilkes. One fragment with Alexander-Yadkin, one with Alleghany, one with Surry.

(4) Stanly. One fragment with Anson-Union, one with Montgomery.

(5) Lee. One fragment with Chatham, one with Randolph.

(6) Moore. One fragment with Randolph, one with Richmond, one with Lee.

(7) Bladen. One fragment with Robeson, one with Sampson.

(8) Pender. One fragment with New Hanover, one with Duplin-Jones.

(9) Granville. One fragment with Person, one with Vance-Warren.

Eight of the extra cuts are in smaller counties (entitled to less than one representative). Splitting such counties is arguably worse since it will be difficult to be elected from the fragments since many fellow county residents will not be able to vote for a candidate, and the fragments will have less influence over the election of "their" representative. Five of the fragments are relatively small (Wilkes(2), Stanly, Pender, and Granville).

In addition, More will not have one whole district and a fragment, but will have three fragments.

Number of groups per number of counties (ie there are 13 2-county groups)
1: 12
2: 13
3: 6
4: 6
5: 4

Number of groups per number of districts (ie there are 11 2-district groups)
1: 13
2: 11
3: 4
4: 6
5: 4
6: 1
11: 1
12: 1

30 counties are in 10 single-member, multi-county groups. All such groups/districts are consistent with the North Carolina constitution. There are an additional 3 single-member, single-county groups.

[jimrtex]

One claim in the state litigation was that the legislative plans violated the Whole County Provisions of the North Carolina Constitution. This was rejected by the North Carolina district court (see page 49 of decision).

Dickson v Rucho opinion by North Carolina district court (PDF)

The court ruled that the legislature had better followed the dictates of Stephenson I and Stephenson II in drawing the maximum number of 2-county groups before then drawing the maximum number of three-county groups, etc. They make note of the deposition of Peterson where it was his opinion that the North Carolina Supreme Court was wrong. The district court said it was bound by the precedent of the North Carolina Supreme Court.

The North Carolina Stephenson guidelines produce a paradox.

Imagine that that the most groups possible was 41, and under Plan A, it was possible to have 12 one-county groups, and 28 two-county groups, encompassing 68 counties. The remaining group would have 32 counties, 9 of which would be cut, which means that this mega group had at least 10 representatives. If the intent was to complete chop up one area with the state with almost total disregard for county boundaries, this would be the way to do it.

Stephenson also ignores that any whole county group electing a single representative is completely consistent with the North Carolina Constitution. If it is possible to arrange a nine-county area into two groups of three and six counties, or four and five counties, the constitution is agnostic as which is better. Here again Stephenson producing a paradox. By giving a preference for a group of three, it also gives a preference for a group of six.

The North Carolina Supreme court affirmed the lower court decision (see page 43, etc. of the PDF)

Dickson v Rucho-1 North Carolina Supreme Court (PDF)

The North Carolina Supreme Court is preparing to consider Dickson v Rucho yet again. After the Alabama decision, the SCOTUS remanded the case to the North Carolina Supreme Court, which confirmed its earlier decision. After the SCOTUS upheld the federal court decisions, they vacated the North Carolina Supreme Court decision. Since the SCOTUS has now said that the North Carolina Supreme Court was wrong, that court may revisit the Stephenson guidelines which says that you first draw VRA districts, and then draw the whole county groups around them. But that was seen as evidence of a predominance of race.

[jimrtex]

This is an alternative cover with 42 whole county groups, that combines Mecklenburg and Union in a two-county group, which permits rearrangement of the groups in south central North Carolina.



Green Unchanged, no overturned districts: 25 groups with 62 districts.

Blue New Groups, no overturned districts: 1 group with one district.

White New groups, overturned districts: 10 groups with 33 districts.

Yellow Unchanged, overturned districts: 6 groups with 30 districts.

Overall, up to 58 (48.3%) districts may have to be redrawn. If we discount the six single-district groups that are in new groups, up to 52 (43.3%) districts may have to be redrawn.

Standard deviation assuming perfect splits of multi-district groups: 2.64%.

Forced county cuts (8 total):

(1) Haywood. One fragment with Jackson-Swain, one with Madison-Yancey.

(2,3) Wilkes. One fragment with Alexander-Yadkin, one with Alleghany, one with Surry.

(4) Stanly. One fragment with Anson-Richmond, one with Montgomery (there are alternative splits possible in this group).

(5) Lee. One fragment with Chatham, one with Harnett.

(6) Robeson. One fragment with Bladen, one with Columbus, and one district wholly within Robeson (there are alternative divisions).

(7) Wayne. One fragment with Sampson, one with Duplin, and one district wholly within Wayne (there are alternative divisions).

(8) Granville. One fragment with Person, one with Vance-Warren.

Number of groups per number of counties (ie there are 16 2-county groups)
1: 11
2: 16
3: 6
4: 6
5: 3

While eliminating (from my 41-group plan) one single-county group, the plan also eliminates one five-county group, and replaces them with three two-county groups.

Number of groups per number of districts (ie there are 11 2-district groups)
1: 14
2: 11
3: 6
4: 5
5: 3
6: 1
11: 1
14: 1

Compared to my 41-group plan, this eliminates one 4-district and one 5-district group, and creates one additional 1-district group, and two additional 3-district groups. It also converts the 12-district Mecklenburg group to a a 14-district Mecklenburg-Union group.

32 counties are in 11 single-member, multi-county groups. All such groups/districts are consistent with the North Carolina constitution. There are an additional 3 single-member, single-county groups.

[jimrtex]

These are the whole county groups which are unchanged from the 2011 law and which do not contain VRA districts. These districts do not need to be modified. The groups contain a total of 53 districts, 45 Republican and 8 Democrat.

These are unchanged county groups that contain overturned districts. New maps can be drawn for each group and slipped in the statewide map with no ripple effects.

This is an alternative cover with 42 whole county groups, that combines Mecklenburg and Union in a two-county group, which permits rearrangement of the groups in south central North Carolina.



Green Unchanged, no overturned districts: 25 groups with 62 districts.

Blue New Groups, no overturned districts: 1 group with one district.

White New groups, overturned districts: 10 groups with 33 districts.

Yellow Unchanged, overturned districts: 6 groups with 30 districts.

So the only seemingly keepable groups from the existing plan (those from the first two maps in this post I'm making) that aren't kept are the Meklenburg group and the adjacent Union-Anson group. Plus one could argue that the group around the Albermarle Sound (which loses Currituck and gains Washington before being divided into two single-district groups; unchanged from your original plan that ended up having one too many districts) could have been kept, but perhaps not while avoiding a county split within and the Currituck-Washington trade makes sense to me.  The blue group in your third map includes territory from the gigantic 20-county group that included three overturned districts.

The Moore-Randolph and Chatham-Lee-Harnett (in green on the 3rd map) groups are the same as the existing plan.

The 20-county snake goes from Stanly to Dare, but by a circuitous route. In the 3rd map, look at the white area. Take, Mecklenburg, Union, and Anson out and the snake begins Stanly-Montgomery-Richmond, and then goes south of Cumberland (Fayetteville). It leaves out Pitt and Wilson which formed an improbable 2-county group (they have a 3-mile wide boundary that has two districts crossing through it, as well as using the middle mile as a boundary jog between the two districts, and continues east to include Pamlico, Hyde, and Dare in the blue group, plus Washington (but not Currituck).

The 9 counties in the northeast in my two groups, were a single two-representative group (with Washington out and Currituck in). Taking Currituck out is what made my split possible. That is only possible by including Pamlico. While the Pamlico-Hyde connection is a tenuous water connection it makes possible the northeastern split, which I think is a feature of my plan, since it produces a third compact whole-county majority BVAP district.

I like that your new groups with more than one district tend to be "non-stringy" (three counties that all border each other or four counties with all but one pair sharing a border, or a four-corners situation - am I right that that is the case with Stanly, Montgomery, Anson and Richmond?).  I think I prefer this plan to your other plan (that has the correct number of districts).

Yes, I noticed that when I realized there were choices when you decide which counties need to have the extra split.

That is a four-corners. The western boundary of Richmond and Montgomery is the center of the Pee Dee River, which including tributaries form county boundaries up to between Yadkin and Surry.

The Montgomery-Richmond boundary is a straight line that crosses the river, where it meets the center-line of the Rocky River which forms the boundary between Stanly and Anson, and Stanly and Union. The straight line forms the boundary between Stanly and Anson, from the center-line of the Pee Dee River to the west bank (about 300 feet).

If you look at smaller scales, it almost looks like Montgomery comes to a point just touching Anson, but there is actually an almost 90 degree intersection of a straight line and the center-line of the river.

One thing I have not mentioned, is that if one is willing to go outside 5% limits slightly, three cuts can be eliminated.

Rowan-Davidson-Stanly-Montgomery is another four corners.

Pitt is 5.8% above two districts. If it were separated from Beaufort-Craven it would eliminate a tiny sliver of Pitt that is attached to those two counties. Those two counties would be -4.8% under.

Johnston is 6.3% above two districts. If it was separated from Duplin-Sampson-Wayne it would eliminate a tiny sliver that is attached to those three counties. The three remaining counties would be 2.6% over.

Surry is -7.3% below a district. If if was separated from Alexander-Alleghany-Wilkes-Yadkin it would eliminate one of the splits of Wilkes. While Wilkes is split twice, about 93% of the county population is in one district (Alleghany-Wilkes could form a two-county group, but a strip is needed to connect Alexander and Yadkin, and a tiny bit is needed to get Surry up to a whole district.

[jimrtex]

The two legislative redistricting committees held a joint hearing on Wednesday. I could not find any audio or video, but there were news accounts, which were somewhat superficial (the editor for the news probably said, "Don't we have a map of North Carolina? Yeah, one with districts would be good. The House and Senate districts are too small. Let's go with the congressional map? Hmm that doesn't look too bad. What about one with the old I-95 district that the Democrats drew?

There are now committee websites. If you click on the 7/26 folder, you will find some maps.

House Select Committee on Redistricting website

Senate Redistricting Committee website

[jimrtex]

Thomas Hofeller drew the current maps, and will likely draw the new maps. He has been involved in redistricting in various states since the 1970s. His testimony basically sealed the case against the current maps.

In his explanation about the process he said that he did not attend any hearings, or committee meetings, or read any transcripts. He then backtracked and said that he might have stuck his head in, but had no reason to stay and listen.

As part of the current litigation he presented what he claims is the optimum plan for whole county groups, and it is from this map that the number of counties that must be redrawn comes from.

Hofeller Optimum Plans (PDF), you should be able to rotate this in your browser or (PDF reader such as Adobe).

Here is the map presented in a form comparable to that for the other plans (the color scheme was adopted from Hofeller's affidavit, with the addition of blue for groups that would require drawing new districts in areas where none were overturned.



His strict reading of Stephenson is that the maximum number of one-county groups must be drawn, followed by the maximum number of two-county groups, etc. His plan does achieve that, with 12 one-county groups, and 17 two-county groups. But as a consequence it ends up three larger monster groups: (1) 6-county groups with 4 representatives on the northern boundary; (2) 6-county, 6 representative snake from Davie to Richmond; and (3) the 7-county, 7-district, 3-legged group  southeast of Raleigh.

A consequence of forming the maximum number of groups is to produce leftover regions which can not be divided into small groups. An extreme case is the nine counties in northeastern part of the state which are split 3 and 6, rather the 4 and 5 in my plan.

If we imagine the drawing of a map much as a master stone mason would carefully choose stones to assemble a wall, the Hofeller plan is the equivalent of an apprentice slapping together smaller stones and then shooting massive amounts of caulk or concrete or plastic to fix his mistakes.

Green Unchanged, no overturned districts: 20 groups with 39 districts.

Blue New Groups, no overturned districts: 4 groups with 12 districts.

White New groups, overturned districts: 11 groups with 32 districts.

Yellow Unchanged, overturned districts: 6 groups with 37 districts.

Overall, up to 81 (67.5%) districts may have to be redrawn. If we discount the five single-district groups that are in new groups, up to 76 (63.3%) districts may have to be redrawn.

Standard deviation assuming perfect splits of multi-district groups: 3.29%.

The deviation for the three monster groups are +4.58% (4 representatives, northern), +3.34% (6 representatives, snake), +5.00% (+4.996%) (7 representative 3-legged monster). Because they are large districts in terms of representatives, they produce an overall imbalance. While single member groups may have large individual deviations, they tend not to accumulate. In the Hofeller map, the thirteen single-member groups have an average deviation of -0.41%. Collectively, they are only 4268 persons short of the population needed for 13 representatives.

The three monster groups have a population equivalent to 17.733 representatives, but are only apportioned 17 representatives. They are cheated out of almost a whole representative.

Forced county cuts (9 total):

(1) Haywood. One fragment with Jackson-Swain, one with Madison-Yancey.

(2) Wilkes. One fragment with Alexander, one with Alleghany-Surry.

(3) Surry. One fragment with Wilkes-Alleghany, one with Stokes-Rockingham.

(4) Rowan. One fragment with Davie, one with Cabarrus-Stanly, one wholly within Rowan.

(5) Stanly. One fragment with Cabarrus-Rowan, one with Montgomery-Richmond.

(6) Harnett. One fragment with Lee, one with Sampson-Johnston-Wayne, one entirely in Harnett.

(7) Sampson. One fragment with Bladen, one with Harnett-Johnston-Wayne.

(8) Granville. One fragment with Person, one with Vance-Warren.

(9) Columbus. One fragment with Robeson, one with Pender.

Number of groups per number of counties (ie there are 17 2-county groups)
1: 12
2: 17
3: 4
4: 3
5: 1
6: 3
7: 1

Number of groups per number of districts (ie there are 11 2-district groups)
1: 13
2: 11
3: 6
4: 5
5: 1
6: 2
7: 1
11: 1
12: 1

30 counties are in 10 single-member, multi-county groups. All such groups/districts are consistent with the North Carolina constitution. There are an additional 3 single-member, single-county groups.

[jimrtex]

The NC legislature is being taken to court (again!) about their slow-walking their constitutional responsibilities. It will be interesting to see what sanctions come down, if any.

https://www.apnews.com/0553c981d6094ac99e9cd2b5859b07cb

Your characterization is not accurate.

The plaintiffs did not bother to file suit until 2015!. They may not have been able to find lawyers, or their lawyers may have been busy with other cases. There was similar litigation in the state courts, which upheld the maps drawn by the legislature.

After the Alabama decision, the SCOTUS accepted the state case, and then remanded it back to the North Carolina Court. The North Carolina Supreme Court once again uphold the state plan, which was then appealed to the SCOTUS, which did not take the case.

Eventually, in August of 2016, the federal district court decided the case. This was obviously too proximate to the November 2016 election to draw districts, especially given the fact that the federal court was concerned about how the line-drawing had been in 2011.

The state appealed the federal case to the SCOTUS. The intemperate federal court ordered North Carolina to draw new districts and hold a special election in 2017. This would require representatives who had been elected to be tossed from office in the middle of their terms, which is an extraordinary solution.

The state asked for a stay of this redistricting order, while the underlying case was being considered by the SCOTUS. At this point the decisions of the federal court and state courts were in conflict, and the SCOTUS would be the appropriate body to resolve the difference. In addition, there was the issue as to whether the plaintiff parties in the state and federal cases were independent of each other (they had the same lawyers, which is perhaps why the federal plaintiffs did not file suit until 2015). The SCOTUS granted the stay.

The state offered to do expedited briefing, which the plaintiffs did not, but the SCOTUS did not accept the offer in any case. The SCOTUS decided the federal case in May - but in their decision they vacated the order to immediately draw districts and hold a special election. They spent half of their opinion explaining why, suggesting that the district court really did not understand their past decisions - that the district court had been injudicious.

The state had to explain to the federal district court that SCOTUS decisions are not final until a certified opinion is issued by the SCOTUS, about a month after the opinion is made.

The official reason for having the hearing today was to help the federal court determine the course of action after their original decision had been tossed by the SCOTUS.

[jimrtex]

The Stephenson decisions by the North Carolina Supreme Court in the 2000s cycle were an attempt by the North Carolina Supreme Court to harmonize the Whole County Provisions of the North Carolina Constitution; the One Man, One Vote decisions of the 1960s and 1970s; decisions of the SCOTUS which mandate single-member districts when at-large elections may submerge racial (or possibly political) minorities in a larger electorate; and provisions of the Voting Rights Act which may mandate disregard for traditional districting criteria such as respect for political subdivisions, including counties.

The role of the North Carolina Supreme Court should be to set objective criteria that a harmonized plan should meet. Instead, they set out a procedure that they thought would meet the criteria. It is analogous to the situation where referees were to determine if a play resulted in a touchdown, but instead of checking for infractions such as multiple forward passes, ineligible receivers, holding, etc., they actually drew up a play to score a touchdown. If the play failed, they might feel obligated to thrust both arms into the air, since it was their play. Or if a team ran their own play, they might throw a flag for illegal procedure.

The 1868 North Carolina Constitution provided that each county should have one representative regardless of population, and that the additional representatives should be apportioned on the basis of population. But with 100 counties, and 120 representatives, there were only 20 extra representatives to be apportioned to the larger counties. Counties with much less than 1/120 of the population would have one representative and counties with several times 1/120 might only be apportioned two representatives.

If the United States House of Representatives had 60 members apportioned among the States on the basis of population, California would have 4 representatives; Texas would have 3; New York, Florida, Illinois, Pennsylvania, and Ohio two each; and all of the remaining States, whether Michigan or Wyoming, one a piece.

A similar provision in Alabama, which required one representative for each county was at the crux of Reynolds v Sims. Alabama had compounded the problem by never reapportioning as was required by the Alabama constitution. Alabama has not since amended its constitution, they just ignore it.

The 1971 North Carolina Constitution removed the requirement that each county have a representative, but included a provision that prevents division of a county. Thus some districts will be multi-county with a single representative, others will be single-county with multiple representatives, and some might be multi-county with multiple representatives.

This map complies with the North Carolina Constitution.



Counties with a population greater than 1/120 of the population would be single-county districts, apportioned additional representatives based on population. Counties with less than 1/120 of the population are aggregated into five contiguous districts: (1) Eastern, which wraps around from Caswell to Columbus, 33 counties with 14 representatives; (2) Western, which stretches from Stokes to Cherokee, 24 counties with 10 representative; (3) Southern, between Charlotte and Fayetteville, 6 counties with 3 representatives; (4) Chatham-Lee, two counties with two representatives; and (5) Lincoln one county with one representative. In total, there are 66 small counties, with a population equivalent to 30 representatives. The 1971 constitution freed up 36 additional representatives, added to the 20 available from the 1868 constitution, to distribute a total of 56 representatives to the 34 largest counties. These 34 counties with 3/4 of the state population are apportioned 90 representatives, or 3/4 of the total.

Of course, there is no reason not to group the smaller counties into single-district groups. There is one three-county two-district group of Alleghany-Stokes-Surry. Otherwise, Stokes would have been a single-county district with 58.7% of a quota.

[jimrtex]

I figured out how to fix my original plan with 43-groups and 121 representatives. Apportion one less representative to Mecklenburg.



The North Carolina Constitution requires that the 120 representatives be apportioned among the districts such that the number of persons per representative is a near as equal as may be. When you apportion 120 representatives using the harmonic means for the divisors, Mecklenburg gets 11 representatives. If there were 121 representatives, Mecklenburg would get it.

The North Carolina Stephenson decision says the deviation should be between -5% and +5% in order to comply with SCOTUS OMOV decisions. But that is not precise. The SCOTUS has sanctioned deviations that go outside that range, particularly in order to recognize political subdivisions such as counties.

Since the North Carolina Constitution forbids division of counties, it is reasonable to go outside the 5% limits to avoid splitting counties. The above map sets Pitt and Surry counties off as single-county districts.

Surry has a deviation of -7.3%. Placing Surry in its own group, eliminates a a cut of Wilkes simply to get Surry into range. The remaining 4-county group of Alexander-Alleghany-Wilkes-Yadkin has two nrepresentatives add a deviation of -1.9%.

Pitt has a deviation of +5.8% for two representatives. Placing Pitt in its own group eliminates a tiny part of Pitt being attached to adjacent counties. The remaining group of Craven and Beaufort has a deviation of -4.8% for two representatives.

Mecklenburg with 11 representatives has a deviation of +5.2%.

Green Unchanged, no overturned districts: 23 groups with 50 districts.

Blue New Groups, no overturned districts: 5 groups with 9 districts.

White New groups, overturned districts: 10 groups with 20 districts.

Yellow Unchanged, overturned districts: 7 groups with 41 districts.

Overall, up to 70 (58.3%) districts may have to be redrawn. If we discount the seven single-district groups that are in new groups, up to 63 (52.5%) districts may have to be redrawn.

Standard deviation assuming perfect splits of multi-district groups: 3.22%.

Forced county cuts (7 total):

(1) Haywood. One fragment with Jackson-Swain, one with Madison-Yancey.

(2) Wilkes. One fragment with Alexander-Yadkin, one with Alleghany. Almost all of Wilkes (95%+) will be with Alleghany.

(3) Montgomery. One fragment with Stanly, one with Randolph.

(4) Robeson, One fragment with Hoke, one with Columbus, and one whole district in Robeson.

(5) Harnett. One fragment with Lee, one with Johnston-Wayne, one entirely in Harnett.

(6) Sampson. One fragment with Bladen, one with Duplin.

(7) Granville. One fragment with Person, one with Vance-Warren.

Number of groups per number of counties (ie there are 17 2-county groups)
1: 14
2: 17
3: 6
4: 6
5: 2

Number of groups per number of districts (ie there are 15 2-district groups)
1: 15
2: 15
3: 5
4: 4
5: 2
6: 2
11: 2

32 counties are in 11 single-member, multi-county groups. All such groups/districts are consistent with the North Carolina constitution. There are an additional 4 single-member, single-county groups.

[jimrtex]

The NC legislature is being taken to court (again!) about their slow-walking their constitutional responsibilities. It will be interesting to see what sanctions come down, if any.

https://www.apnews.com/0553c981d6094ac99e9cd2b5859b07cb

The federal district court rejected the plaintiffs demand for a special election. Some of the districts will have new boundaries for the 2018 election. Given that 2018 is an off year in North Carolina with no presidential, no gubernatorial, and no senatorial elections, Republicans are likely to gain legislative seats.

[jimrtex]

You must be logged in to read this quote.

I disagree with that and have quite a few reasons to support that:
- It's a Trump Midterm and right now Democrats are far more energized than Republicans which is visible in Special Elections all around the country. Turnout drop of Dems will likely be smaller then in past Midterms.
- Money. Gov Cooper efforts have already cashed in three times as much the Reps have got so far. Eric Holder's group is also very interested in investments in NC
- Redistricting. The current map is a racial gerrymander against black voters. Therefore a new constitutional map has to give African Americans more voting power.
- Voting Patterns. There a quite a lot Republican Legislators in districts Clinton carried. So these Republicans could go under if these Democratic trends continue downballot

Edit: And some of the districts will be modified in 2018? Even Republicans admit that they have to alter two thirds of all General Assembly districts to correct the 28 unconstitutional ones.

Do you have presidential election results by House District? What is your source for the claim of "quite a few".

Democrats House member hold five McCain 2008 districts, while Republicans hold one Obama 2008 districts. I compared county results from 2008 to 2016 (remember Obama carried North Carolina in 2008), just to make sure that whiter areas in western North Carolina had not voted against Obama because he was black. But support for Clinton dropped off the table in those areas. They didn't vote for Clinton because she was a Democrat and because she was Clinton. Some of those McCain districts that are now held by Democrats, might have been carried by Clinton, since some are in more urban areas.

Her support held up much better in urban areas, but the Democrats already hold most of the seats in those areas:

Mecklenburg (Charlotte) 8:4
Wake (Raleigh) 8:3
Cumberland (Fayetteville) 3:1
Durham-Orange (Durham-Chapel Hill) 5:0
Guilford (Greensboro-High Point) 3:3

Democrats gained a few districts in 2016 which might have been drawn as Republican opportunity districts. The districts were not really drawn in order to bleach adjacent Republican districts. They were drawn to meet an arbitrary quota, to get Eric Holder's USDOJ to preclear the map.

You can not assign persons voters on the basis of race. "Vote here", "Why?", "You're Black" is illegal.

"Vote here", Why?", "You're Black" is just as illegal if done to make their votes more efficient.

You can assign voters on the basis of where they live. That is the whole point of districts, no? If there is a community, it has to be based on locality. Areas where there are AME churches, and barber shops and beauty salons that specialize in treating persons with particular hair types, and style preferences, tend to be predominately black. The churches and barber shops are part of what defines the community.

In the more rural areas of the state, the percentage of blacks goes down as you go south. Bladen-Sampson-Duplin is 29% BVAP and the population for two representatives. Just by chance you might get two 29% BVAP districts, which might well be two Republican districts. There might be some differentiation, so it is 34% and 24%, so it is an even split.

But if you start shifting people around you can quickly dislodge the few remaining Democrats in the area.

Voters who are energized by opposition to Trump tend to be concentrated in areas that would vote for Maduro. In 2018, there will be no presidential election, no gubernatorial election, no senatorial election, no council of state election. Turnout will be down, just like it is in every off-year election.

[jimrtex]

No, just no.  With how gerrymandered these districts were, they will not be gaining seats.  That's partisans dreams.

Jim knows that, he's just trolling.

The districts are not as gerrymandered as some fantasize.

In Cumberland, there were two districts that were 47% and 54% BVAP after the 2010 census. They are now both just over 50%. The problem is that they were achieve a numeric target to get preclearance from the USDOJ. Though Eric Holder might want to say that a plan retrogresses because it is not a Democratic gerrymander, he knew that would not stand up in coutt.

You might remember the Ohio redistricting contest, where the NAACP was one of the sponsors. One of their criteria was to draw two black majority senate districts in Cuyahoga County. The only way to do that is to have two districts running east to west, ignoring ward lines, and placing two very black house districts with one quite white house district in western Cleveland. You were literally having to choose between putting Parma or Lakewood in a black district, because at 7% black, they were blacker than the alternatives. Because I couldn't do this, my plan was not even scored.

They also wanted a majority black district in the Cleveland-Akron area. I don't know why they didn't like my solution of using the median of interstates to connect the two cities. I even included the parts of the cloverleafs between the interstates so that you could drive between Cleveland and Akron and stay in the district.

[jimrtex]

No, just no.  With how gerrymandered these districts were, they will not be gaining seats.  That's partisans dreams.

Jim knows that, he's just trolling.

The districts are not as gerrymandered as some fantasize.

Having super-majorities in a state that votes 3% Republican is not normal.   No matter how much you try to spin it, the maps help Republicans massively.

The population distribution in North Carolina is not normal (in a statistical sense). The mean, which you have cited, is not representative.

57.3% of voters were in precincts where Trump received a majority of the two-way vote. If each precinct elected its own representative and the representative had a weight vote, Trump supporters would would have 57.3% of the vote. As you aggregate such vote into larger units, they tend to be more likely to support Trump.

[jimrtex]

No, just no.  With how gerrymandered these districts were, they will not be gaining seats.  That's partisans dreams.

Jim knows that, he's just trolling.

The districts are not as gerrymandered as some fantasize.

The courts disagree with you, which I know really bothers you.

I may have an advantage over you in that I have read the actual opinions and have done considerable independent analysis.

[jimrtex]

You must be logged in to read this quote.

I disagree with that and have quite a few reasons to support that:
- It's a Trump Midterm and right now Democrats are far more energized than Republicans which is visible in Special Elections all around the country. Turnout drop of Dems will likely be smaller then in past Midterms.
- Money. Gov Cooper efforts have already cashed in three times as much the Reps have got so far. Eric Holder's group is also very interested in investments in NC
- Redistricting. The current map is a racial gerrymander against black voters. Therefore a new constitutional map has to give African Americans more voting power.
- Voting Patterns. There a quite a lot Republican Legislators in districts Clinton carried. So these Republicans could go under if these Democratic trends continue downballot

Edit: And some of the districts will be modified in 2018? Even Republicans admit that they have to alter two thirds of all General Assembly districts to correct the 28 unconstitutional ones.

Do you have presidential election results by House District? What is your source for the claim of "quite a few".

You talked about "state legislative gains". So before I made this statement I checked the State Senate first. According to Daily Kos there a four GOP senators in Clinton districts and no Democratic senators in Trump districts.

Today I looked up the House numbers:
Trump/Dem are: HD-22, HD 47, HD 66, HD 115, HD 116 (5)
Clinton/Rep are: HD-36, HD 104, HD 105 (3)

So here you are right that there are Republican pick-up opportunities as well. HD-22 has a typical white southern Democrat as its representative, but in the other four districts HRC had at least 44%. It going to be interesting how the new map will change that.

Thank you.

Brisson (HD-22) is quite interesting. He has 4 consecutive elections that he has won with a 52:48 margin. In 2012 and 2016 the results were in the Democratic primary, and he was unopposed in the general. In 2010 and 2014 he defeated a Republican. He is walking a fine line between being too conservative for Democrats, and being too Democratic for Republicans. HD-45 was 45% for Obama, and you suggest that it is even less for Clinton. Brisson is quite vulnerable to redistricting, since he could get an entirely new set of voters.

HD-47 is a 51% NAVAP (Lumbee) district in Robeson County. The district was 58% Obama, so there must have been a huge dropoff. I suspect that Lumbee politics are in an entirely different dimension than national or state HDpolitics. Despite a 51% NAVAP, only 44% of registrants are NA. This might be due to less participation, or how persons are classified by the Census and the Board of Elections.

In 2011 party registration
Whites: 58% D, 22% R, 20% N
Blacks: 87% D, 3% R, 10% N
Native Am: 80% D, 8% R, 12% N

HD-66 is the white district, entangled with HD-48. This makes the incumbent less vulnerable to a challenge by a black Democrat. This district is also vulnerable to a challenge.

HD-115 and HD-116 are in Buncombe (Asheville). They were narrowly for McCain 49.3% and 45.7%. They may have been part of an attempt to draw two R seats in Buncombe.

HD-36 and HD-104 and HD-105 are in Wake (Raleigh) and Mecklenburg (Charlotte). The Republicans lost other seats in those counties in 2016, so if they wish, they can shore up the remaining seats.

[jimrtex]

No, just no.  With how gerrymandered these districts were, they will not be gaining seats.  That's partisans dreams.

Jim knows that, he's just trolling.

The districts are not as gerrymandered as some fantasize.

Having super-majorities in a state that votes 3% Republican is not normal.   No matter how much you try to spin it, the maps help Republicans massively.

The population distribution in North Carolina is not normal (in a statistical sense). The mean, which you have cited, is not representative.

57.3% of voters were in precincts where Trump received a majority of the two-way vote. If each precinct elected its own representative and the representative had a weight vote, Trump supporters would would have 57.3% of the vote. As you aggregate such vote into larger units, they tend to be more likely to support Trump.

That's a complete nonsense argument considering precincts don't have anywhere near equal population.   Some have a few hundred (or few dozen...) while others have tens of thousands.   

The kind of legislature that would produce would be nightmarishly unequal.

You misunderstood what I was saying.

Alamance 1, had 2276 votes, so its representative would have 2276 votes. Yancey 11 PRI, had 837 votes, so its representative would have 837 votes, and similarly for the 3207 other precincts in between.

The total number of votes cast in precincts carried by Trump was 2,661,852.
The total number of votes cast in precincts carried by Clinton was 1,890,095

Trump carried precincts with 58.5% of the vote.

I made a mistake in my previous message and read the wrong row of my spreadsheet. Trump had 51% or more of the vote in precincts with 57.2% of the total vote.

You may have understood me to say that Trump had carried 57.2% of the precincts. In fact, he carried 63.4% of precincts. But that would be a nonsensical argument to make, and I would not make nonsensical arguments.

Do you know how to use a spreadsheet? I'll show you how to reproduce my results:

(1) Go to the North Carolina State Board of Elections website.
(2) Go to the Election Results and then the November 2016 Election..
(3) Click on "Downoads" in the upper left.
(4) Click on Precinct Election Results and download the zip file.
(5) Click on (layout) to see format of the file.
(6) Extract the txt file from the zip file and load it into excel.
(7) The txt file has one row for each candidate for each precinct for each race, so you will want to get rid of the non-presidential races and the votes for other than Trump and Clinton.

Let me know if you (or anyone else) gets this far.

[jimrtex]

This is an alternative map with 44 groups, that keeps 12 districts in Mecklenburg, converts, Pitt, Johnston, and Surry to single-county groups slightly outside the 5% deviation range, and keeps the Randolph-Moore, and Chatham-Lee-Harnett groups from the existing map. It focuses the map drawing on the areas that contain the overturned districts.



Green Unchanged, no overturned districts: 22 groups with 49 districts.

Blue New Groups, no overturned districts: 6 groups with 13 districts.

White New groups, overturned districts: 9 groups with 16 districts.

Yellow Unchanged, overturned districts: 7 groups with 42 districts.

Overall, up to 71 (59.2%) districts may have to be redrawn. If we discount the seven single-district groups that are in new groups, up to 64 (33.3%) districts may have to be redrawn. The changes to the Alleghany-Alexander-Wilkes-Yadkin, Union-Anson-Stanly-Montgomery, and Johnston groups are quite minimal, just enough to balance populations. Excluding those 8 districts, only 56 districts (46.7%) need to be redrawn.

Standard deviation assuming perfect splits of multi-district groups: 2.95%.

Forced county cuts (8 total):

(1) Haywood. One fragment with Jackson-Swain, one with Madison-Yancey.

(2) Wilkes. One fragment with Alleghany, one with Alexander-Yadkin.

(3) Stanly. One fragment with Union-Anson, one with Montgomery.

(4) Lee. One fragment with Chatham, one with Harnett.

(5) Robeson. One with Hoke, one with Columbus, one entirely in Robeson.

(6) Pender. One fragment with Bladen, one with Duplin.

(7) Wayne. One fragment with Sampson, one with Duplin-Pender, one entirely in Wayne.

There are alternative arrangements for 6 and 7. The above puts the two smallest counties, Bladen and Pender in a district, with a small piece of Pender attached to Duplin. It then provides districts based in the two medium sized counties of Sampson and Duplin and tops them off with the surplus from Wayne. Wayne keeps a whole district, the alternative tend to chop either Sampson or Duplin in half.

(8) Granville. One fragment with Person, one with Vance-Warren.

Placing Pitt and Johnston in single county groups, eliminates gratuitous chops.

Number of groups per number of counties (ie there are 14 2-county groups)
1: 15
2: 14
3: 6
4: 6
5: 3

Number of groups per number of districts (ie there are 14 2-district groups)
1: 14
2: 14
3: 5
4: 6
5: 2
6: 1
11: 1
12: 1

30 counties are in 10 single-member, multi-county groups. All such groups/districts are consistent with the North Carolina constitution. There are an additional 3 single-member, single-county groups.