North Carolina Legislative Redistricting

[jimrtex]

The North Carolina Constitution requires that legislative districts not divide counties. This violates the OMOV provisions of the US Constitution to the extent that it would require districts with population outside the 5% range.

The North Carolina has harmonized the requirements of the North Carolina and US Constitutions by requiring that:

(1) The legislature draw VRA districts.
(2) The legislature draw non-VRA districts wholly within a single county where possible (e.g. identify the single counties that can have whole districts drawn within).
(3) Identify the maximum number of two-county groupings that the same can be done.
(4) Identify the maximum number of three-county groupings and so on.
(5) In drawing districts within multi-county groupings boundary crossing should be minimized.

The federal district court appears to disagree with the approach of creating the VRA districts before identifying the multi-county groupings, or at least to the extent that a different set of multi-county groupings were created after the VRA seats were identified suggests race sorting.

Anyhow, this is a map showing the number of representatives per county, and identifies the 12 counties which can support a whole number of districts within 95%-105% of the quota.

The House has 120 members and the Senate has 50, so there is no nesting, and the two houses are treated the same but independently.



Collectively, the 12 counties have a population equivalent to 47.030 members, and independently round to 47 members, so it looks OK to give Mecklenburg 12 districts and Wake 11 despite the small difference.

In trying to identify two-county groupings it looks like there are lot of traps where some reasonable two-county groupings might prevent creation of other groupings. North Carolina is relatively narrow in a north-south direction. Where it fattens out in the east, there are relatively few people.

[jimrtex]

Here is map dividing the state into whole county districts, groups of counties with a population equivalent to a whole number of representatives.



In the western part of the state, these are the same as currently in use, and the districts within will not need to be redrawn. The exceptions are in Mecklenburg (Charlotte) and Guilford (Greensboro, High Point) where 6 districts were invalidated. In these two counties, new maps will be redrawn. In Gulford is likely that all districts will be modified, but there will be no partisan change. In Mecklenburg, the changes might strengthen a couple of Democratic wins, in districts which must have been meant for a Republican to win, but were won by a Democrat in 2016.

In the eastern part of the state many of the county groups are quite different. This appears to be because the VRA districts were drawn first, and then county groups drawn around them based on the fragments left over after the VRA districts were drawn. The VRA districts in the more rural parts of the east are quite ugly, miniature versions of the old NC-1. It is possible that the the Gingles conditions do not exist because the black population is not compact.

[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.



Forsyth County (Winston-Salem) is not included among the challenged areas because it was impossible to create two majority black districts, which left the legislature to do a Democratic pack.

[jimrtex]

These are new whole county groups, that likely do not have VRA consequences, so that districts may be drawn within the areas without worrying about race.



Currently, Randolph is paired with Moore, and there are districts in the northern part of Randolph and southern part of Moore that can presumably be maintained without change. The third district that is mostly in southern Randolph could be extended in northern Montgomery, with little political change and the incumbent being re-elected.

Currently Stanly and Montgomery are tied in a whole county group that loops around Cumberland (Fayetteville) and extends to Dare, that can not be explained other than by a predominance of race. Currently there is a district that includes Stanly and the northern 2/3 of Montgomery, that could be modified to be Stanly and souther 2/3 of Montgomery that will re-elect an incumbent.

Currently, Chatham is in a district that extends into Lee. The Lee portion of the district would be replaced with the northern portion of Moore. Chatham is a moderately strong Democratic county, and is only 13% black, and the district would likely continue to be Democratic. The current incumbent is black, and lives in Lee County. But politicians are not a protected class. This is one consequence of the federal district court's drive to hold special elections in 2017 - it is dependent on truncating terms of representatives elected in 2016.

[jimrtex]

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.



Mecklenburg (Charlotte) has 12 districts, 5 of which, 99, 101, 102, 106, and 107, are majority Black VAP, but only three, 99, 102, and 107 were challenged and overturned. It appears that areas in North Charlotte were divvied up between 102 and 107 in particular to push their Black percentage up. The logical way to fix 107 is to lop off its southern tail, and distribute that between 101 and 106. This will simply result in more variation among the Black VAP% but not change any results.

It is conceivable that the Republicans could make a minor change to 88 that would trigger a special election in 2017, that might gain them a seat narrowly lost in 2016.

It appears that Charlotte is a mini-Atlanta, attracting talented or entrepreneurial blacks from rural areas where they can move into the middle class, and not being as tied to the mills and tobacco as areas further north? Perhaps Charlotte is more attractive to black South Carolinians who might find Greenville or Spartanburg more hostile, Columbia relatively moribund, and Charleston charming, but not as much for service workers.

**********

Wake (Raleigh) has 11 districts, 2 of which, 33 and 38, are majority Black, and were overturned. The next highest black share was 26% BVAP. 33 and 38 are not the most convoluted districts by far. It would be relatively easy to tweak 40 and 49, which were narrow Democratic victories in 2016, to force a special election.

Though Wake has almost as many districts as Mecklenburg (the difference is due to rounding), it has much fewer majority Black districts. Perhaps the jobs open to Blacks are clerical or support jobs in the state government or at NC State, steady and with pensions, but not particularly high paying?

***********

Guilford (Greensboro-High Point) has 6 districts, 3 of which, 57, 58, and 60, are majority Black, and were overturned. Simply unsplitting precincts is unlikely to change any results. It might be possible to create a High Point district that would be competitive.

***********

Durham (Durham) and Orange (Chapel Hill) have 5 districts, three in Durham, one in Orange, and one straddling the county line. Two districts in Durham, 29 and 31, are majority black and were overturned. All five districts are Democratic-held. Moving districts around isn't going to change matters much.

**********

Cumberland (Fayetteville/Fort Bragg) has four districts, two of which, 42 and 43 are majority black and were overturned. Changing boundaries is unlikely to have much effect. It might lock in a district that a Democratic narrowly won in 2016.

*********

Person, Granville, Vance, and Warren have two districts, one of which, 32 is majority black and was overturned. There is a strong racial and political gradient from west to east:

Person 26% BVAP, 57% Trump
Granville 33% BVAP, 50% Trump.
Vance 48% BVAP, 37% Trump
Warren 51% BVAP, 33% Trump

If we assume 100% Black support for Clinton,  and equal turnout then white Trump support was around 70% or higher. There is racially polarized voting.

Warren and Vance are short of a quota, and need about 1/4 of Granville to complete the district. The division of Granville is quite intricate. While Granville overall is 33% black, the portion added to district 32 is 54% black, and that to District 2 is 26% Black.

If the division of Granville was not racially differentiated, then the BVAP% for HD-2 would increase from 26% to around 30%, while that for HD-32 would drop from 50% to 46%. A compact area including Oxford could be placed in HD-32 and would bump that 46% up a bit, and that of HD-2 down a tad. There will be no political consequences.

**********

Franklin and Nash have two districts, one of which HD-7 is overturned. Franklin is 27% Black, lower in the suburban areas adjacent to Wake in the south, higher in the rural north. Nash is 41% black, again lower in suburban areas to the southwest, and higher in the north and around Rocky Mount which is on the eastern edge of the county.

Eric Holder's DOJ blocked Rocky Mount switching to nonpartisan election on the belief that the community would not know who their candidate of choice was unless he had a '(D)' next to his name. The eastern portion of Rocky Mount, particularly the portion in Edgecombe is more black than the west, and the area to the city's west is the whitest part of Nash.

The current map differentiated the population of the counties so that HD-7 has a BVAP% of 50%, and HD-25 has a BVAP of 16%. The division in Franklin is HD-7 44%, HD-25 17%; and in Nash HD-7 52%, HD-25 15%. The boundary between the two districts is quite intricate as the Nash portion of HD-7 is just barely above a majority BVAP% by avoiding white areas, and overall at 50% by selecting black areas in Franklin that are significantly black.

A district entirely in Nash centered on Rocky Mount would be around 39% BVAP, leaving HD-25 around 26% BVAP. The North Carolina Supreme Court's interpretation of the North Carolina constitution favors fewer boundary crossings.

[jimrtex]

These are new county groupings, with the overturned districts highlighted.

[Kevinstat]

Franklin and Nash have two districts, one of which HD-7 is overturned. Franklin is 27% Black, lower in the suburban areas adjacent to Wake in the south, higher in the rural north. Nash is 41% black, again lower in suburban areas to the southwest, and higher in the north and around Rocky Mount which is on the eastern edge of the county.
...
The current map differentiated the population of the counties so that HD-7 has a BVAP% of 50%, and HD-25 has a BVAP of 16%. The division in Franklin is HD-7 44%, HD-25 17%; and in Nash HD-7 52%, HD-25 15%. The boundary between the two districts is quite intricate as the Nash portion of HD-7 is just barely above a majority BVAP% by avoiding white areas, and overall at 50% by selecting black areas in Franklin that are significantly black.

A district entirely in Nash centered on Rocky Mount would be around 39% BVAP, leaving HD-25 around 26% BVAP. The North Carolina Supreme Court's interpretation of the North Carolina constitution favors fewer boundary crossings.

The 39% BVAP population for the district entirely in Nash is less than the 41% in the county as a whole.  I understand that the eastern part of Rocky Mount is very white, but you also said the southeast of Nash County was quite white and that latter area could probably be connected to Franklin County without doing anything grotesque.  That could potentially push the BVAP of the district entirely in Nash up a point or two from the 41%, or at least not push it down two points.

[Kevinstat]

Franklin and Nash have two districts, one of which HD-7 is overturned. Franklin is 27% Black, lower in the suburban areas adjacent to Wake in the south, higher in the rural north. Nash is 41% black, again lower in suburban areas to the southwest, and higher in the north and around Rocky Mount which is on the eastern edge of the county.
...
The current map differentiated the population of the counties so that HD-7 has a BVAP% of 50%, and HD-25 has a BVAP of 16%. The division in Franklin is HD-7 44%, HD-25 17%; and in Nash HD-7 52%, HD-25 15%. The boundary between the two districts is quite intricate as the Nash portion of HD-7 is just barely above a majority BVAP% by avoiding white areas, and overall at 50% by selecting black areas in Franklin that are significantly black.

A district entirely in Nash centered on Rocky Mount would be around 39% BVAP, leaving HD-25 around 26% BVAP. The North Carolina Supreme Court's interpretation of the North Carolina constitution favors fewer boundary crossings.

The 39% BVAP population for the district entirely in Nash is less than the 41% in the county as a whole.  I understand that the eastern part of Rocky Mount is very white, but you also said the southeast of Nash County was quite white and that latter area could probably be connected to Franklin County without doing anything grotesque.  That could potentially push the BVAP of the district entirely in Nash up a point or two from the 41%, or at least not push it down two points.

I just noticed that the Republican State Rep. in that conglomerate is from Rocky Mount, and the Democrat (a black woman) is from Franklin County.  If the Republican General Assembly gets to try and fix the court's issues with their plan, then they'll want to help their incumbent as much as possible, while the black Democrat is probably just as screwed in a 26% BVAP district as she would be in a district with a lower BVAP.

[jimrtex]

Franklin and Nash have two districts, one of which HD-7 is overturned. Franklin is 27% Black, lower in the suburban areas adjacent to Wake in the south, higher in the rural north. Nash is 41% black, again lower in suburban areas to the southwest, and higher in the north and around Rocky Mount which is on the eastern edge of the county.
...
The current map differentiated the population of the counties so that HD-7 has a BVAP% of 50%, and HD-25 has a BVAP of 16%. The division in Franklin is HD-7 44%, HD-25 17%; and in Nash HD-7 52%, HD-25 15%. The boundary between the two districts is quite intricate as the Nash portion of HD-7 is just barely above a majority BVAP% by avoiding white areas, and overall at 50% by selecting black areas in Franklin that are significantly black.

A district entirely in Nash centered on Rocky Mount would be around 39% BVAP, leaving HD-25 around 26% BVAP. The North Carolina Supreme Court's interpretation of the North Carolina constitution favors fewer boundary crossings.

The 39% BVAP population for the district entirely in Nash is less than the 41% in the county as a whole.  I understand that the eastern western part of Rocky Mount is very white, but you also said the southeastwest of Nash County was quite white and that latter area could probably be connected to Franklin County without doing anything grotesque.  That could potentially push the BVAP of the district entirely in Nash up a point or two from the 41%, or at least not push it down two points.

Franklin is to the west, Nash to the east. Wake (Raleigh) is the blue county to the south. Rocky Mount is on the Nash-Edgecombe county line (Edgecombe is east of Nash). The Edgecombe portion is about 30% of the city and is 85.5% BVAP. The Nash portion is 70% of the city and 45.5% black. The portion of Rocky Mount closer to the eastern county line is more black, than areas further west or even outside of the city. Growth is to the west. I-95 also passes through that area. Rocky Mount is 60 miles from Raleigh, so you wouldn't pick Rocky Mount if you took a job in Raleigh. You might commute if you were from Rocky Mount, or had a spouse with a job in Rocky Mount.

The areas closer to Raleigh are around 10-15% black. This might be a residual population, as suburban areas are developed in a formerly rural area, or perhaps it is accessible to middle class blacks in 2010. The area west of Rocky Mount is whiter than that. The population has been divided quite carefully. Most precincts in the county were split, except in the extreme southeast. This division of precincts was a factor in the district court's decision.

What I did was place all of Franklin in HD-25, and then started filling in the area to the west of Rocky Mount and the spots along the Franklin-Nash county line.

[Kevinstat]

Franklin and Nash have two districts, one of which HD-7 is overturned. Franklin is 27% Black, lower in the suburban areas adjacent to Wake in the south, higher in the rural north. Nash is 41% black, again lower in suburban areas to the southwest, and higher in the north and around Rocky Mount which is on the eastern edge of the county.
...
The current map differentiated the population of the counties so that HD-7 has a BVAP% of 50%, and HD-25 has a BVAP of 16%. The division in Franklin is HD-7 44%, HD-25 17%; and in Nash HD-7 52%, HD-25 15%. The boundary between the two districts is quite intricate as the Nash portion of HD-7 is just barely above a majority BVAP% by avoiding white areas, and overall at 50% by selecting black areas in Franklin that are significantly black.

A district entirely in Nash centered on Rocky Mount would be around 39% BVAP, leaving HD-25 around 26% BVAP. The North Carolina Supreme Court's interpretation of the North Carolina constitution favors fewer boundary crossings.

The 39% BVAP population for the district entirely in Nash is less than the 41% in the county as a whole.  I understand that the eastern western part of Rocky Mount is very white, but you also said the southeastwest of Nash County was quite white and that latter area could probably be connected to Franklin County without doing anything grotesque.  That could potentially push the BVAP of the district entirely in Nash up a point or two from the 41%, or at least not push it down two points.

Franklin is to the west, Nash to the east. Wake (Raleigh) is the blue county to the south. Rocky Mount is on the Nash-Edgecombe county line (Edgecombe is east of Nash). The Edgecombe portion is about 30% of the city and is 85.5% BVAP. The Nash portion is 70% of the city and 45.5% black. The portion of Rocky Mount closer to the eastern county line is more black, than areas further west or even outside of the city. Growth is to the west. I-95 also passes through that area. Rocky Mount is 60 miles from Raleigh, so you wouldn't pick Rocky Mount if you took a job in Raleigh. You might commute if you were from Rocky Mount, or had a spouse with a job in Rocky Mount.

The areas closer to Raleigh are around 10-15% black. This might be a residual population, as suburban areas are developed in a formerly rural area, or perhaps it is accessible to middle class blacks in 2010. The area west of Rocky Mount is whiter than that. The population has been divided quite carefully. Most precincts in the county were split, except in the extreme southeastwest. This division of precincts was a factor in the district court's decision.

What I did was place all of Franklin in HD-25, and then started filling in the area to the west of Rocky Mount and the spots along the Franklin-Nash county line.

I was thinking west and southwest there even though I wrote east and southeast.  Your enhanced description was helpful though.  I think one thing that threw me off is that your black percentages for Franklin and Nash counties at the beginning of that segment were not necessarily of BVAP.  I would imagine the district entirely in Nash that you describe would have a higher BVAP than Nash as a whole, if you're putting the white suburban area west of Rocky Mount in the Franklin district.  It sounds like the part of Nash along the border with Franklin has both blacker and whiter parts.

[jimrtex]

Franklin and Nash have two districts, one of which HD-7 is overturned. Franklin is 27% Black, lower in the suburban areas adjacent to Wake in the south, higher in the rural north. Nash is 41% black, again lower in suburban areas to the southwest, and higher in the north and around Rocky Mount which is on the eastern edge of the county.
...
The current map differentiated the population of the counties so that HD-7 has a BVAP% of 50%, and HD-25 has a BVAP of 16%. The division in Franklin is HD-7 44%, HD-25 17%; and in Nash HD-7 52%, HD-25 15%. The boundary between the two districts is quite intricate as the Nash portion of HD-7 is just barely above a majority BVAP% by avoiding white areas, and overall at 50% by selecting black areas in Franklin that are significantly black.

A district entirely in Nash centered on Rocky Mount would be around 39% BVAP, leaving HD-25 around 26% BVAP. The North Carolina Supreme Court's interpretation of the North Carolina constitution favors fewer boundary crossings.

The 39% BVAP population for the district entirely in Nash is less than the 41% in the county as a whole.  I understand that the eastern part of Rocky Mount is very white, but you also said the southeast of Nash County was quite white and that latter area could probably be connected to Franklin County without doing anything grotesque.  That could potentially push the BVAP of the district entirely in Nash up a point or two from the 41%, or at least not push it down two points.

I just noticed that the Republican State Rep. in that conglomerate is from Rocky Mount, and the Democrat (a black woman) is from Franklin County.  If the Republican General Assembly gets to try and fix the court's issues with their plan, then they'll want to help their incumbent as much as possible, while the black Democrat is probably just as screwed in a 26% BVAP district as she would be in a district with a lower BVAP.

The Republican was elected in 2010, when HD-25 was entirely in Nash County, with part of Rocky Mount excised out. So part of the reason that HD-25 runs up into western Rocky Mount was incumbent protection.

The Democrat was appointed to her position (in North Carolina, vacancies are filled by political parties). The previous representative, Angela Bryant, moved to the senate after the death of a senator. A party appointment might be made of someone who didn't necessarily have a local political base to win in an open election.

Before 2011, HD-49 was all of Franklin plus an arm that ran up to Roanoke Rapids. HD-7 was most of Halifax, with an arm down into Nash to take in part of Rocky Mount. The incumbent in HD-7 in 2010 lives in Rocky Mount, and was re-elected in 2012 when the district swapped out Halifax for more of Nash and northern Franklin. After Bryant moved to the senate, her replacement was from newly added Franklin County.

A feature of county-based redistricting rules is that it often works like term-limits. But it is biased against more rural counties, since you can maintain incumbent districts in urban counties since the lines aren't restricted, particularly when you don't have formal subdivisions.

Conceivably, Collins(the R) could be elected in a 39% BVAP district, since he formerly represented more of Nash County, lives in Rocky Mount, and it would be a special election in 2017. A problem with a special election in 2017 is that it requires terms being truncated to one year. This is biased against black incumbents since they represent many of the districts being redrawn.

[jimrtex]

This covers all the single-district whole county groups in far eastern North Carolina. Such districts and groups are wholly compatible with the North Carolina constitution. Particular attention is paid to the overturned HD-5 in the northeast corner.



HD-5 is comprised of three whole counties, plus a piece of Pasquotank. As is fairly typical, the black districts are drawn into larger towns, not to take the whole town, but only the blacker areas, so that not only are counties split, so are towns. In this case the salient is into Elizabeth City. HD-5 is 53.4% BVAP. HD-5 is represented by a black Democrat from Hertford.

HD-1 is the remainder of the area and is 18.6% BVAP. It includes Currituck, which is part of the Hampton Roads metropolitan area and is only 5.9% BVAP. HD-1 is represented by a white Republican from Chowan.

**********

The (red) county group of Hertford, Bertie, Chowan, Washington, and Tyrrell is 51.3% BVAP is compact and generally south of Albermarle Sound or west of the Chowan River estuary.

**********

The (blue) county group of Gates, Perquimans, Pasquotank, and Camden is 31.3% BVAP and is north of Albermarle Sound. It was about 55:45 McCrory:Cooper, so would likely elected a Republican, even in an open election.

*********

HD-6 includes Dare county, and is 16.7% BVAP. In 2014, it narrowly elected a white Democrat, who in 2015 became an independent, and caucused with the Republicans. He didn't seek re-election in 2016, and a Republican was narrowly elected. Dare County was about 5% more favorable for Trump and McCrory than it was for the Republican representative candidate. It was the Beaufort portion of the district that won the election for the Republican, who is from Dare County.

The green group of counties drops Washington and Beaufort, and gains Currituck and Pamlico, with the BVAP dropping to 9.0%, and should be a Republican seat going forward.

**********

Halifax and Northampton form HD-27. It has a 53.2% BVAP. It is currently represented by a white Democrat from Northampton. He was first elected in 2004, advancing to the 2nd primary in 2004 with 28% of the vote, and then becoming the nominee with 51.4% of the vote, in part due to low turnout. Since the 2011 redistricting, he has been challenged in the 2012, 2014, and 2016 primaries, with his support declining from 62%, to 58%, to 52%. He has never had a Republican opponent, but did have a Libertarian in one election.

**********

Edgecombe and Martin form HD-23. It has a 51.4% BVAP and is currently represented by a black Democrat from Edgecombe (Rocky Mount).

**********

Cartaret and Jones form HD-13. It has a 9.1% BVAP and is represented by a white Republican from Carteret.

[jimrtex]

What I did was place all of Franklin in HD-25, and then started filling in the area to the west of Rocky Mount and the spots along the Franklin-Nash county line.

I was thinking west and southwest there even though I wrote east and southeast.  Your enhanced description was helpful though.  I think one thing that threw me off is that your black percentages for Franklin and Nash counties at the beginning of that segment were not necessarily of BVAP.  I would imagine the district entirely in Nash that you describe would have a higher BVAP than Nash as a whole, if you're putting the white suburban area west of Rocky Mount in the Franklin district.  It sounds like the part of Nash along the border with Franklin has both blacker and whiter parts.

I've started reading through the district court's decision in Covington, which begins with a general overview, and then goes through a district by district analysis.

The state's map-drawer noted that except for the VRA, the double traversal of the county line by HD-7 and HD-25 would violate the the NC Supreme Court's harmonization of the state constitution with OMOV. That is, division of Nash is necessary because it is entitled to 1.209 representatives, but it is not necessary to divide Franklin.

It is conceivable that a double traversal might comply with the VRA. It is the stated target of BVAP% > 50%, the splitting of voting precincts, communities (Rocky Mount), and the irregular shape of the district that are the main problems.

[jimrtex]

The next area is divided into three county groups, and contains two overturned districts:




The groups and counties:

Blue:

Pitt 2.116 quotas, 31.8% BVAP
Beaufort 0.601 quotas, 24.4% BVAP
Craven 1.303 quotas, 21.2% BVAP

Green:

Wilson: 1.022 quotas, 37.7% BVAP

Red:

Greene: 0.269 quotas, 37.0% BVAP
Lenoir: 0.749 quotas, 39.0% BVAP


The current districts include:

HD-24: the majority-black district in Wilson and Pitt counties, that has been overturned. It has a BVAP of 56.6% and connects the town of Wilson to Greenville. Its current representative is Jean Farmer-Butterfield from Wilson. She is the former wife of congressman G.K.Butterfield (NC-1). They were divorced in 1991. She began serving in the legislature in 2003, he in Congress in 2004. Prior to that he was a judge, including an appointed stint on the North Carolina Supreme Court.

HD-8: is the remainder of Wilson, plus the portion of Pitt west of HD-24. It has a BVAP of 27.3%. Its representative, a white Republican, is also from Wilson, and was first elected in 2012 after redistricting, she squeaked through with a 162-vote win in 2016.

The boundary between Wilson and Pitt counties is about 3 miles long. HD-24 passes through the northeastern part, about one mile wide; HD-8 passes through the southwestern part, about one mile wide; and the middle part forms a jog in the boundary between the two.

HD-9: is the eastern portion of Pitt and is 18.2% BVAP. Its representative is a white Republican from Pitt (Greenville).

HD-12: is the black-majority district in Greene, Lenoir, and Craven counties, that has been overturned. It is 50.0% BVAP. The portion of the district that stretches the length of Craven, splits 22 of the 23 precincts in the county. The boundary is particularly convoluted around New Bern (where the 4 is). Its representative is a black Democrat from Lenoir (Kinston)

HD-10: wraps around the HD-12 in Lenoir and Greene counties, and extends east of HD-12 in Craven down to as far as New Bern. It also has an arm down the western side of Wayne county. It is 14.9% BVAP. Its representative is a white Republican from western Wayne (Goldsboro) who is now the House majority leader. It almost looks like the district was drawn to include him, but he was first elected in 2012 after narrowly winning the primary. It is possible that he moved into the district to run. If a seat is created for him, it will be in the next county group to the west,

HD-3: includes the portions of Craven east of the Neuse River (towards Beaufort) and in the boot that are not in HD-12, along with Pamlico, and a portion of Beaufort, generally south of the Tar River (estuary). It is 19.0% BVAP. Its representative is a white Republican from Craven (New Bern).

HD-6: includes the remainder of Beaufort, generally north of the Tar River. It shifts eastward, and is covered above.


Conceptual districts:

Green:

Wilson would form a district that has a BVAP of 37.7%. The incumbents from HD-8 and HD-24 both live in Wilson. They were elected from districts that were 27% and 57% BVAP, and would have to compete in a district that was 38% BVAP. Clinton, Ross, and Cooper carried the county in the 2016, presidential, senatorial, and gubernatorial races, by about 5%.

The timing of the next election may be critical. In 2018, North Carolina will not have a presidential, gubernatorial, or senatorial election. This would ordinarily favor Republicans based on turnout. Turnout for a 2017 special election restricted to some General Assembly seats is an interesting question. Ordinarily low turnout would favor Republicans, but in this case, Blacks might turn out, particularly if they are convinced that the Republicans got rid of the majority-minority districts and called a special election.

Red:

Lenoir and Greene would form a district that has a BVAP of 38.4%. Greene is quite rural, the county seat of Snow Hill only has 1600 people. Kinston, the county seat of Lenoir, has about the same number of people. The incumbent for HD-12 is from Kinston, but the district was carried by Trump, Burr, and McCrory.

Blue:

Pitt, Craven, and Beaufort will be divided among four districts. Craven and Beaufort together barely have enough for two districts (-4.8% too small, but within the 5% limit). But Pitt has too much for two districts (+5.8% too large). Within the overall group, the deviation is 0.5%.

Based on minimum county traversals, two district would be wholly within Pitt, one wholly within Craven, and one comprised of Beaufort, a portion of Craven, and a small part of the Pitt.

The Pitt portion of HD-8 and HD-24 combined is 46.0% BVAP. HD-24 in Pitt is 54.0% BVAP, and HD-8 is 33.6% BVAP. That is the two areas are black and less-black, rather than black and white. HD-9 the only district wholly within Pitt is 18.3%. The boundary between HD-9 and the other two districts is reasonably clean. An oddity is that it appears the boundary is more regular when it follows roads and physical features, than if it followed precinct boundaries.

The Pitt portion of HD-8 and HD-24 have slightly more population than needed for a district, which means that a slight adjustment would need to be made to balance population. A small area along the border (around 8,000 persons or less) need to be detached for adding to the Beaufort-Craven-Pitt district.

The new district in western Pitt would be an open seat, and would presumably be won by a black Democrat, probably from Greenville. HD-9 is represented by a white Republican from Greenville, and the boundary would be drawn so as to keep him in the district.

The district in Craven would logically be formed in the boot, and include New Bern, Cherry Point, and Havelock. Craven overall is only 21.2% BVAP, and I doubt that this district would be much above that. The incumbent for HD-3 is a white Republican from New Bern and would presumably be elected from this district.

The final district would be 60% in Beaufort (the whole county), 30% in Craven, including the area east of the Neuse River, and 10% into Pitt. I suspect that the district would have a BVAP in the 25-30% range. There is no incumbent in the district, but the winner in an open race would probably be a white Republican.


Summary:

Currently, there are four (white) Republican and two (black) Democratic representatives, though a white Democrat came close in in 2016.

The new configuration would elected one black Democrat, along with a 50/50 shot at each of two others. The other three seats would be won by white Republicans.

[jimrtex]

The next group of four counties is strung together because I could not find smaller alternative groupings.



Only Wayne has a large enough black population to have VRA implications.

Wayne: 1.543 quotas, 30.8% BVAP
Johnston: 2.125 quotas, 14.9% BVAP
Harnett 1.443 quotas, 20.3% BVAP
Lee 0.728 quotas, 19.4% BVAP

Harnett should have one whole district, Johnston two, and Wayne one.

The other districts would be Lee and part of Harnett; and part of Harnett, a small part of Johnston, and part of Wayne.


Current districts:

HD-21 is the black majority district that begins in Wayne and then heads down into Duplin and Sampson. It is 51.2% BVAP. About 3/5 of the district is in Wayne, concentrated in the Goldsboro area. The Wayne portion of the district is 53.2% BVAP. Its representative is a black Democrat from Sampson (Clinton).

HD-4 wraps around HD-21 on the west, north, and east, and then takes in the whiter areas of Duplin. It is 15.8% BVAP. Its representative is a white Republican from Duplin (Warsaw).

HD-10 (see above) includes the western part of Wayne, and then snakes into Greene, Lenoir, and Craven. About 1/3 of the district is in Wayne, including its representative, who is also the majority leader, who has a Goldsboro address, but appears to live west of the city.

HD-26 and HD-28 are in Johnston, and appear to be Raleigh suburbs. They are 16.3% and 13.8% BVAP, respectively, and are represented by white Republicans from Clayton and Pine Level.

HD-22 takes in a small tab of southern Johnston, necessary since Johnston is 6.2% more than needed for two districts.

HD-53 is the eastern part of Harnett, and has a 21.5% BVAP. Its representative is a white Republican from Harnett (Dunn), who happens to be the House redistricting committee chair and a named defendant. Dunn is in the extreme southeastern corner of Harnett.

HD-51 is remainder of Harnett (about 1/4 of the county) and most of Lee. It is 15.1% BVAP and represented by a white Republican from Lee (Sanford).

HD-54 is a small intrusion into Lee, plus all of Chatham. It is 17.6% BVAP. It is represented by a black Democrat from Lee (Sanford). The Lee portion is 35.9% BVAP, but relatively small in population. The inclusion appears to be an attempt to pack Democrats into Chatham, which is fairly solidly Democrat despite a small black population.


Starting from the west, the districts can be largely maintained.

District A: All of Lee, and the western part of Harnett. This would pair two incumbents from Lee, which is not unreasonable in that Lee has less than a quota.

District B: The central part of Harnett. While this maintains much of HD-53, it would no longer have an incumbent.

District C: and District D: These would be largely the same as HD-26 and HD-28 in Johnston. They would contract somewhat northward for reasons of population balance.

District E: This would be placed in Wayne and include the Wayne portion of HD-21, and parts of HD-4 to the east, with the intent of maximizing the BVAP population, which will be somewhere in the mid- to high-30s%. It does not have an incumbent.

District F: This would use the remainder of Wayne, Johnston, and Harnett. This would be an ugly seat since Wayne and Harnett will be connected by a narrow strip across the southern tip of Johnston. It would also pair two incumbents, including the House majority leader and the chair of the redistricting committee.

An alternative would be to split HD-28 in Johnston, and join the two parts with Harnett and Wayne. This would produce more compact districts, but fragments Johnston more than necessary. Arguably the number of county line traversals remains the same. This might be possible to be done in a way that avoids pairing incumbents.

[jimrtex]

The next group is Duplin, Sampson, and Bladen, three rural counties in southeastern North Carolina.




Duplin is 26.0% BVAP, and 0.736 districts
Sampson is 27.2% BVAP and 0.798 quotas
Bladen is 34.2% BVAP and 0.443 quotas

Overall the area is 28.3% BVAP


Current districts:

HD-21: is a black-majority district that has been overturned. The portion that is in Duplin and Sampson is 48.2% BVAP and only constitutes about 2/5 of a district. Its representative is a black Democrat from Sampson (Clinton). Clinton is near the protrusion towards the '2' in Sampson. Clinton with about 8500 persons is the largest town in the three-county area.

HD-4: is the eastern part of Duplin, and then wraps around HD-21 on three sides in Wayne. Slightly less than 1/2 the district is in Duplin. The portion of the district in Duplin is 14.9% BVAP. Its representative is a white Republican from Duplin (Warsaw).

HD-22: is the western part of Sampson, most of Bladen, and a smidgen of Johnston. It is 26.4% BVAP. Its representative is a conservative Democrat from Bladen (Dublin). He has had narrow wins in each of the last four elections. Two were in the primary, in 2012 and 2016 over opponents who thought he sided with Republicans too often. In those years he had no Republican opponent. In 2010 and 2014 he defeated a Republican.

HD-46: is a small part of southeastern Bladen (about 1/7 of the the county).


Prospective Districts:

The Whole County Provision requires minimum traversals of county lines when creating districts. Therefore, Sampson will be divided between the two districts, with roughly 2/3 of the Sampson population with Bladen, and 1/3 with Duplin. An attempt should be made to keep all of Clinton in one district.

It does not appear possible to get much racial differentiation, so both districts will be within a couple of percentage points of the overall 28.3% BVAP, so the division should be based on good government principles of compactness, and not splitting towns or precincts.

A north-south division of Sampson might result in a more compact division , since this would make the district that includes Bladen more compact, but this should not be done if it requires a division of Clinton.

[jimrtex]

These groups are along the South Carolina line. There is a significant Native American (Lumbee) population in the area, particularly in Robeson County.




County Populations:

Columbus: 30.0% BVAP, 3.0% NAVAP, 0.731 quotas.
Robeson: 20.0% BVAP, 37.1% NAVAP, 1.688 quotas.
Hoke: 34.1% BVAP, 9.2% NAVAP, 0.591 quotas.
Scotland: 36.9% BVAP, 9.8% NAVAP, 0.455 quotas.
Richmond: 29.2% BVAP, 2.3% NAVAP, 0.586 quotas.


Current Districts:

HD-46: consists of Columbus, a portion of southern Robeson, and a small part of Bladen. It has a 25.4% BVAP, and 5.8% NAVAP. Columbus is about 3/4 of the district. The portion in Robeson is not too convoluted, as an attempt is to made to avoid black and Native American population, which it largely succeeds at. The Robeson portion of HD-46 has a 13.4% BVAP and 16.1% NAVAP, both of which are below the county average. Its representative is a white Republican from Columbus (Tabor City). The district has been competitive. The district was drastically reconfigured in 2011, from a district that was in Robeson, Hoke, and Scotland, to a district that was 3/4 in Columbus. The southern tip of the 2010 district, became the northern tip of the 2012 district, and provided an opportunity for a white Democrat from Columbus to be elected. He was re-elected in 2014, defeating the current incumbent, but did not run again in 2016. An altercation between the Republican and Democrat before the 2016 election made national news. Assault charges filed by both men were dismissed after the election.

HD-47: is a Native American majority district in Robeson. It is 51.0% NAVAP, 18.4% BVAP. It is quite compact except for a tentacle of HD-48 that tries to reach concentrations of blacks in Lumberton. HD-47 is a VRA protected district. It is currently represented by a Lumbee Democrat from Lumberton, who defeated a 10-term Lumbee Democrat in the 2010 primary.

HD-48: is a black majority district in Hoke, Richmond, Robeson, and Scotland counties, that has been overturned. It is 50.7% BVAP, and 11.1% NAVAP. Roughly 1/4 of the district is in each of the counties, that is composed of tentacles and corridors connecting them. It is decidedly not compact. It is represented by a 7-term black Democrat from Scotland (Wagram). His last opponent was in the general election in 2010 which he won with 75% of the vote against a Republican opponent.

HD-66: is a district in Hoke, Richmond, Robeson, Scotland, and a small part of Montgomery counties. It is 21.9% BVAP and 5.3% NAVAP. It can basically be described as the parts of Hoke, Richmond, and Scotland that aren't in HD-48. It is represented by a white Democrat from Richmond (Rockingham). He was first elected in a dramatically different configuration of HD-66, winning the general election by 61%-39% over a Republican opponent. He has had no opponents in the primary or general election since the 2011 redistricting.


Prospective districts:

Richmond and Scotland. 32.6% BVAP, and 5.5% NAVAP. It would pair the incumbents from HD-48 and HD-66, both Democrats. The district would have been a Trump-Cooper district. Overall, it appears that rural districts in this part of the state continue to be Democratic at the legislative level. In 2011, registration in the district would have been 64.2% Democrat, 16.6% Republican. In 2017 it is 56.3% Democrat, 18.0% Republican. Registration is a lagging indicator, otherwise a 3:1 registration advantage would not translate into a tossup at the presidential and gubernatorial level.

To minimize county line traversals, the Hoke-Robeson-Columbus districts would be:

(1) Hoke and about 1/4 of Robeson
(2) About 60% of Robeson
(3) Columbia and about 1/6 of Robeson

(3) is similar to HD-46, with the loss of a small portion of Bladen and a bit more of Robeson. The current boundary attempts to avoid black and Native American areas of Robeson (the portion of HD-46 in Robeson is 13.4% BVAP and 16.1% NAVAP. The small part of Bladen is 18.4% BVAP. Columbia is 30.0% BVAP, and the inclusion of the small part of Bladen plus Robeson, drew the overall BVAP down to 25.4%. Adjustments should pull the BVAP closer to 30%. It current white Republican incumbent should be able to continue to be elected.

(2) would continue to be entirely in Robeson. It will continue to be a Native American VRA district.

(1) would consist of Hoke and a portion of Robeson. This would push (2) to the southeast, since the only alternative would be to try to incorporate the Lumberton tentacle of HD-48 into this district (1). Hoke is 34.1% BVAP and 9.2% NAVAP. This district would have a slightly lower BVAP, and slightly higher, to significantly higher NAVAP.

Pushing (2) to the southeast will likely reduce the NAVAP to below 50.0% and increase the BVAP above the current 16.9%. Since this is presumably a VRA district, it might be possible to show that the district is still a performing Native American district. Alternatively, the whole county provisions of the state constitution might be ignored, but I'm not sure how.

You could go east from Hoke into Cumberland to get an area with less Native American population. But Cumberland has four districts of its own. So a Cumberland district would have to come into Sampson or Bladen, and push the Sampson-Bladen district into southern Robeson.

[Kevinstat]

These groups are along the South Carolina line. There is a significant Native American (Lumbee) population in the area, particularly in Robeson County.
[...]
County Populations:

Columbus: 30.0% BVAP, 3.0% NAVAP
Robeson: 20.0% BVAP, 37.1% NAVAP
Hoke: 34.1% BVAP, 9.2% NAVAP
Scotland: 36.9% BVAP, 9.8% NAVAP
Richmond: 29.2% BVAP, 2.3% NAVAP

And the quotas of each county are...?

[jimrtex]

These groups are along the South Carolina line. There is a significant Native American (Lumbee) population in the area, particularly in Robeson County.
[...]
County Populations:

Columbus: 30.0% BVAP, 3.0% NAVAP
Robeson: 20.0% BVAP, 37.1% NAVAP
Hoke: 34.1% BVAP, 9.2% NAVAP
Scotland: 36.9% BVAP, 9.8% NAVAP
Richmond: 29.2% BVAP, 2.3% NAVAP

And the quotas of each county are...?

Added. Also, the quotas for all counties are in the first map.

I also added the prospective districts for this area.

[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.

[Kevinstat]

[...]
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:
[...]

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.

[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

[Kevinstat]

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 Bladen with a population of 7.349 quotas (or 1.04986 quotas per district). This will force redistricting down to the block level.

Correct?  With Duplin instead of Bladen, the math is almost as tight (from simple adding of the thousandths of a quota from your map at the beginning of the thread, I get 7.642 or 8 * ~0.955).  With both Duplin and Bladen, the group is actually a pretty nice ~8.085 or 8 * ~1.011.

[The (or Their)] North [Carolina] 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.

Yeah, a better algorithm would be: form maximum number of groups with no "additional county splits" (which would include the single-county groups but also any multi-county single-district groups), form maximum number of groups with only one "additional county split", form maximum number of groups with only two "additional county splits", ...

By an "additional county split", I mean a county with less than 1.05 quotas being divided into two districts, a county with less than 2.1 quotas being divided into three districts, etc.

It's still probably not as good as your proposal to simply limit the number of "additional county splits," but it would be better than what they have come up with.

[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.