Texas 2020 House Apportioment

[jimrtex]

Why can't the texas GOP gerrymander the Beoumont district? Is it VRA protected?

It may have been, particularly under Section 5. It might not be now, because it likely fails the Gingles test. As Texas population increases, district sizes increase.

Jefferson(Beaumont-Port Arthur) and Smith(Tyler) are the only two counties with enough population to be carved into house districts. Smaller counties have to be assembled into districts made up of whole counties. Tyler is a strong Republican area, because of its urban population and the legacy of the East Texas Field. It was electing Republicans when no other areas along the eastern border were electing Democrats. In 2000, there were 13D's and 3 R's from the area. 8 of the D's had no general election opponent.

Many of the D's were WD40's (white Democrats over 40, with their hair slicked back).

Beaumont and Port Arthur have a strong union presence because of the refineries and ship building. Those who made their money in Spindletop have long ago moved on. Beaumont had a significant population. It formed a second tier with cities like Waco, Austin, Corpus Christi, and El Paso (behind Houston, San Antonio, Dallas and Fort Worth). But it has stagnated. The population of Jefferson County is roughly the same as it was in 1960, while that of Texas has tripled.

In the 1970's, Jefferson County had three representatives, and part of another district. The 1970's decision that required election from single-member districts, was because of minority voters being outvoted in at-large elections. So in Jefferson County, you could create a black district. In a smaller county with a 30% black population, and one district, you have a district with a significant black population. If there are three districts, you can create one with at least a majority black population.

Because of the stagnant population, residential segregation patterns have persisted. In a more dynamic economy, middle class blacks can move to new housing in the suburbs. If the population is static, you only have to build enough to account for reduced family sizes (more households for the same population) or replacement housing for low quality houses. Some of the that can be built in the same neighborhoods.

Jefferson County is 34% black. Beaumont is 47% black, Port Arthur 42%, Nederland 4%, Groves 1%, Port Neches 1% (the three Mid-County cities have about 50,000 persons, so they are not insignificant enclaves.

Elsewhere, Democrats were gradually defeated. This was particularly true after redistricting. Because of the requirement of districts being made up whole counties, and a declining population share, the districts are necessarily radically changed. Sometimes two incumbents would be paired, and one would lose the Democratic primary. Or they might have an unfamiliar county added to the district, removing a familiar county.

Bob Johnson (incumbent) won't be the incumbent in the new district. Ballots in Texas don't indicate incumbents. Voters won't recognize the name. If told that he owns an insurance agency, restaurant, etc. over in Hopperville, they might reply, "I haven't been to Hopperville in ages", or remember when the Hopperville Hoppers kept the local Loopertown Leapords out of the state playoffs. Johnson is a stranger to them.

When Democrats were the majority in the Texas House (up until 2002), being a Democrat was an advantage, since it meant better committee assignments.

As time went on, Republicans became dominant at the presidential and state levels. But you could still vote in the Democratic primary where county officers and legislators were chosen. Remember that Rick Perry was a Democratic legislator, and claims he had never met a Republican until he went away to college. He was from west Texas, but for this discussion it is no different. Sometime, all the county commissioners, and other county officials would announce that they would be running as Republicans. Texas does not have party registration, you can choose your party every two years.

In 2004, 2006, and 2008 Chuck Hopson(D) narrowly won re-election, the last time by 120 votes, when he received less than 50% of the vote due to a Libertarian candidate. In 2010, Chuck Hopson(now R) was easily re-elected with 76% of the vote. His district was changed by the 2012 redistricting, adding Nacogdoches which was more than half the new district. He lost in the primary.

In 2010, Republicans won 11 of 13 East Texas seats, including three they finally flipped that year.  Six of the incumbents were unopposed. After the election, one of the two Democrats, Allan Ritter joined the Republican caucus. Along with two other post-election party switchers, Republicans reached their maximum of 101 of 150 House members.

By the 1990's redistricting, Jefferson had been reduced to 2 whole districts and a small fraction of a third. If an area's population is static, while the state's populations is tripling then, the representation for the county will be reduced by 2/3. This is what happened in Jefferson as its representation has decreased from 4 districts to 1-1/3 districts between 1960 and 2020.

Democrats controlled the 1990's redistricting and their aggressive gerrymandering for Congress, the Senate, and House resulted in more minority House districts. In Jefferson, they created a black district, a white district, and the little bit of the Mid County area was added to an inter-county district. At this time, all would be Democratic districts.

The 2000's redistricting was the first to be controlled by Republicans. A Democratic-controlled House, and a Republican-controlled Senate were unable to agree on legislative districts. Under the Texas Constitution, redistricting in that case was done by the Legislative Redistricting Board, which is comprised of statewide office holders, whom were then all Republican except for House Speaker Pete Laney.

Statewide, they simply corrected many of the excesses of the Democratic gerrymander in larger counties. They eliminated districts in rural areas, and added districts in suburbs. At the time, the eliminated rural seats were not necessarily Republican-held, but the new suburban seats would be.

Jefferson was down to about 1.8 districts. Ordinarily, Jefferson would have been joined with Chambers, but Chambers was the only option for Galveston as well. This meant that the Texas Constitution had to be violated in order to comply with OMOV (or for the LRB to discover that Galveston could have been paired with Harris). The LRB divided Orange, which was a smaller county. They added a tentacle into the city of Orange and added that to the black district, and the white district was wholly in Jefferson.

At that time, all the districts in southeast Texas were Democratic-held. So if anything this would make it easier for a Republican to be elected, particularly if there was a retirement. As it was, one district in southeast Texas was eliminated, one flipped in 2002, and another in 2004, so as it was, the southeast Texas delegation went from D6:R1 in 2000, to D3:R3 in 2004. Over the decade, the Republican districts became out of reach for Democrats, and another flipped in 2010.

Allan Ritter was elected without opposition as a Democrat for most of the decade.. After the 2010 election he switched to the Republican Party. There was likely an understanding that he would have a safe seat as a Republican.

By 2010, Jefferson was down to 1.5 districts. The black district was entirely in Jefferson, and the white district took in 1/3 of the county plus Orange. The black district was 67% Obama, while the white district was 24% Obama (in 2012). If you balanced the two, you could have two 48% Obama districts, which might both be Democratic districts.

Allan Ritter was elected as a Republican without opposition in his new district. He retired, and in 2014 a Republican Wade Phelan was elected with 74% of the vote. It is noteworthy that he is from Orange rather than Jefferson like Ritter was.

So the 2010's redistricting was useful to create a Democratic sink, while rewarding a Democratic for flipping a seat, and converting it to a Republcan seat. At the same time, it avoided an additional issue for VRA litigation.

By 2020, Jefferson will be down to 1.3 districts. HD-22 currently has a 51% BCVAP, while HD-21 has an 8% BCVAP. The districts are about as racially differentiated as possible. If HD-22 stays in Jefferson, the district will drop to about 45% BCVAP. This would mean that it fails the Gingles. If you have an area that is 51% BCVAP, and the surroundings areas are 8% BCVAP, there is no way to expand the area and maintain a majority BCVAP.

But a 45% BCVAP district will continue to elect a black Democrat. Blacks will constitute 75% of the Democratic primary vote, and the district will be 60% Democrat.

If you start chopping up the black areas, you might have to violate the state constitution, and also divide cities. You could still a court to rule that you are violating the 15th Amendment.

[jimrtex]

Districts 79 through 82 are in far west Texas



HD-79 (+0.55%) change -0.111 per 5 El Paso districts).  El Paso(31%, 7% of county), Val Verde(25%), Reeves(8%), Pecos(8%), Ward(6%), Brewster(5%), Presidio(3%), Hudspeth(2%), Kimble(2%), Sutton(2%), Crockett(2%), Schleicher(1%), Jeff Davis(1%), Culberson(1%), Menard(1%), Edwards(1%), Terrell(0%), Loving(0%). Cities: Del Rio, Socorro, and San Elizario CDP.

HD-80 (-0.30%) change 0.022. Ector(84%), Andrews(10%), Winkler(4%), Crane(2%). Cities: Odessa, West Odessa, and Andrews.

HD-81 (+0.37%) change 0.075. Midland(89%), Martin(3%), Reagan(2%), Upton(2%), Coke(2%), Irion(1%), Glasscock(1%), Sterling(1%). Cities: Midland, Big Lake, Stanton.

HD-82 (+0.48%) change -0.136. Tom Green(61%), Brown(19%), Runnels(5%), Coleman(4%), McCulloch(4%), San Saba(3%), Mills(2%), Concho(1%),. Cities: San Angelo, Brownwood, and Brady.

El Paso currently has five undersized districts (-4.5% deviation). Once the deviation exceeds 5%, population has to be added to bring the district up to the quota, plus make up any difference due to relative loss of population throughout the decade. In effect, the fifth El Paso district is largely forced out of the county and expands into 17 sparsely populated counties to the east.

Alternatively, the district could be considered as a successor to an existing district that loses Maverick and Kinney, picks up a bit of El Paso and expands into the Hill Country.

HD-80, HD-81, and HD-82 are anchored by the cities of Odessa, Midland, and San Angelo respectively. The anchor counties of Ector, Midland, and Tom Green are too small to be divided, but too large to be merged with another similar county. The overall effect is that the representative will likely be from the anchor city, which has a majority of the population, while each decade the additional counties will get reshuffled, as the districts expand.

Note that Midland and Ector, and some other Permian Basin counties have at least kept up with the statewide growth of 18% for the decade.

34 counties have grown faster than the State as a whole:

Houston area (7) Brazoria, Chambers, Fort Bend, Galveston, Harris, Montgomery, Waller.
DFW area (8) Collin, Denton, Ellis, Hood, Kaufman, Parker, Rockwall, Tarrrant.
San Antonio (5) Bexar, Comal, Guadalupe, Kendall. Wilson.
Austin (4) Bastrop, Hays, Travis, Williamson.

Permian Basin (7) Andrews, Ector, Gaines, Loving, Martin, Midland, Sterling.

Others (3) Brazos (Texas A&M), Frio (Eagle Ford?), Hudspeth (immigrant detention?)

[jimrtex]

Districts 83 through 90 are in west Texas.



HD-83 (+0.88%) Change -0.087. Wise(35%), Erath(22%), Palo Pinto(14%), Eastland(9%), Comanche(7%), Somervell(5%), Jack(4%), Hamilton(4%). Cities: Stephenville, Mineral Wells, and Decatur.

HD-84 (+0.85%) Change -0.146. Taylor(69%), Jones(10%), Nolan(7%), Callahan(7%), Stephens(5%), Shackelford(2%). Cities: Abilene, Sweetwater, and Breckenridge.

HD-85 (+1.67%) Change -0.200. Wichita(66%), Young(9%), Wilbarger(6%), Clay(5%), Archer(4%), Haskell(3%), Hardeman(2%), Knox(2%), Baylor(2%), Throckmorton(1%), Foard(1%). Cities: Wichita Falls, Vernon, and Burkburnett.

HD-86 (+0.47%) Change -0.172 Howard(18%), Gray(11%), Gaines(11%), Scurry(9%), Terry(6%), Dawson(6%), Yoakum(4%), Mitchell(4%), Childress(4%), Garza(3%), Crosby(3%), Lynn(3%), Floyd(3%), Wheeler(3%), Fisher(2%), Donley(2%), Hall(1%), Collingsworth(1%), Dickens(1%), Briscoe(1%), Stonewall(1%), Cottle(1%), Motley(1%), Kent(0%), Borden(0%), King(0%). Cities: Big Spring, Pampa, and Snyder.

In west Texas there are nine counties that have the population equivalent to more than one half a house district: Ector (Odessa), Midland (Midland), Tom Green (San Angelo), Taylor (Abilene), Wichita (Wichita Falls), Lubbock (Lubbock) (2 districts), Potter (Amarillo), and Randall (Amarillo).

Because of their size they must each be placed in separate districts, where they will contain a majority of the district population and serve as anchors. Overall, west Texas has population for 12 districts. This means that there are three unanchored districts in west Texas. As population share is eliminated over the coming decades, these unanchored districts will be eliminated.

HD-86 will likely be gobbled up by the nine surrounding anchored districts in 2030. It may be increasingly difficult to create the anchored districts and avoid nearby districts.

HD-87 (-0.03%) Change -0.162 (for HD-87 and HD-88, -0.081 per two districts). Lubbock(60%, 38% of county), Hale(17%), Hockley(12%), Lamb(7%), Bailey(4%), Cochran(1%). Cities: Lubbock, Plainview, and Levelland.

HD-88 (-0.03%) Lubbock(100%, 62% of county). City: Lubbock.

HD-88 is entirely within the city of Lubbock, with about 1/4 of the city in HD-87. The current district has a strange intrusion from the south that reaches the campus of Texas Tech, and an extrusion that reaches (the former) Reese AFB. These are unlikely to reflect partisan concerns. They are either to preserve separate districts for incumbent representatives, or to give both districts part of the economic engines of the area.

I simplified the boundaries, such that HD-87 is more peripheral areas of Lubbock, which may help the smaller counties be more competitive in determining representations.

HD-89 (+0.51%) Change -0.075. Randall(71%), Deaf Smith(9%), Parmer(5%), Castro(4%), Dallam(4%), Swisher(4%), Hartley(3%), Oldham(1%). Cities: Amarillo, Hereford, and Canyon.

HD-90 (+0.31%) Change -0.191 Potter(61%), Moore(11%), Hutchinson(11%), Ochiltree(5%), Carson(3%), Hansford(3%), Hemphill(2%), Lipscomb(2%), Sherman(2%), Armstrong(1%), Roberts(0%). Cities: Amarillo, Dumas, and Borger.

[Kevinstat]

I see that according to the projected decimal "quotas" you've been using (perhaps different from what you expect to happen in some cases, like in Travis County, but what you're using all the same), Harris County is the one county with a projected quota that is within 5% of more than one integer.  In fact Harris's 24.719 could be (25, -1.12%) (which you use, understandably), (24, +3.00%) or even (26, -4.93%).  Unless, that is, there is some court precedent forbidding one of those last two options (particularly the last one that is more than a whole Representative different from Harris County's "ideal" (fractional) number of State Representatives.  I'd be curious to know what the legal precedent is there.  With Bexar County being projected close to a range which would be within 5% of two integers (the smallest such range, interestingly enough), the ability to either round up or round down could be relevant if it could potentially allow the deviations in single-county districts/conglomerates add up closer to 0 (right now, it's very close but Travis County getting within 5% of 7 and/or Montgomery County ending up more than 5% over 3 could upset that) and thus make the math less tight elsewhere.  Although I guess in that case giving Bexar County an 11th district would make things even worse, but you could think of it as giving Bexar only 10 districts if its quota was 10.494 (4.94% greater than 10; only 4.60% less than 11).  I'd be interested to know about the legal possibilities here.

As you've noted, the last range of decimal numbers that isn't within 5% of any integer is the range from 9.45 (5% greater than 9) to 9.50 (5% less than 10).  Once you get up to 10.45 (5% less than 11; only 4.5% greater than 10), you have at first small but then larger and larger ranges of decimal numbers within 5% of two integers until the last range that isn't within 5% of two integers is the range from 18.90 (5% greater than 18) to 19.00 (5% less than 20).  The lowest range of decimal numbers within 5% of three integers (or the lowest range of decimal numbers that's within 5% of an integer that is a whole number or more off from the number itself) is the range from 20.90 (5% less than 22) to 21.00 (5% greater than 20).  The range of numbers within 5% of 24, 25 and 26 (the range that you project Harris County to be in) is half a "quota" long (from 24.700 to 25.200).

[Kevinstat]

I also meant to ask: are any counties' 2010 decimal quotas (besides Harris's obviously) (the quotas that governed what the current districts could be) in any of the ranges of numbers that are within 5% of two integers?  Like between 10.45 and 10.50 for Tarrant?  Based on what I know that's the only other one possible, as I know your 10.424 projection for Bexar (just below the lowest such range) has it growing faster than the state as a whole and that Dallas's 2010 quota was 14.118 (6.88% below 15).

I'm guessing no for Tarrant, that it was was over 10.50 in 2010 (it was obviously over 10.45) as otherwise it would have been noteworthy that it was apportioned 11 districts in 2010 rather than 10.  Although the harmonic mean of 10 and 11 (the point at which a Tarrant County voter's share of a representative would be the same amont off the statewide average with 10 representatives as 11) is 10.476 and the geometric mean (the number that's the same % {above 10, below 11} that {11, 10} is {above, below} it) is 10.488, so arguably giving Tarrant 11 Representatives in the 2011/2012 redistricting would have been justified as long as Tarrant's quota was above either of those numbers, even if it was below 10.50 and thus in a "range of flexibility."

I see that Harris's 2010 quota was 24.4 (if you round to the nearest 0.1), which is only between 5% of 24 and 25.  Anything between 24.350 and 24.450 is below the arithmetic (24.500), geometric (24.495) and harmonic (24.490) means of 24 and 25 (they get closer together the higher you get), so it's not surprising that Harris County was apportioned only 24 Representatives last time, even if there is some flexibility (I'm not sure if there is or not).

[Kevinstat]

HD-88 is entirely within the city of Lubbock, with about 1/4 of the city in HD-87. The current district has a strange intrusion from the south that reaches the campus of Texas Tech, and an extrusion that reaches (the former) Reese AFB. These are unlikely to reflect partisan concerns. They are either to preserve separate districts for incumbent representatives, or to give both districts part of the economic engines of the area.

I simplified the boundaries, such that HD-87 is more peripheral areas of Lubbock, which may help the smaller counties be more competitive in determining representations.

The real beneficiaries might of that might be people in say, Shallowater or Wolfforth (still in Lubbock County but outside the City of Lubbock itself, and between the City of Lubbock and at least some (all in the case of Shallowater) of the other counties in the 2-district conglomerate).  Of course the current conglomerate may be different, but I'm talking about the benefit of smoothing out the boundaries of the main Lubbock district as opposed to making minimal changes to that district and putting the rest of Lubbock County in with the other counties in your proposed (possibly new) conglomerate.

[jimrtex]

I see that according to the projected decimal "quotas" you've been using (perhaps different from what you expect to happen in some cases, like in Travis County, but what you're using all the same), Harris County is the one county with a projected quota that is within 5% of more than one integer. 

I'll address the estimate issue in this post, and the apportionment issue in the second.

I have used the July 2017 Census estimate to (exponentially) project to April 2020. That is what I based the apportionment on.

Travis has a population of 1,030,522 in the 2010 Census and was entitled to 6.111 representatives. The July 2017 Census estimate is 1,226,698, which would entitle Travis to 6.501 representatives. That represents a 2.5% (compounded) rate of increase.

Projected to April 2020, that would result in 1,313,448 persons, which is equivalent to 6.647 representatives. It appears that growth has slowed some in Travis, and looking at annual increases I would expect it to be just short of 1.3 million. That may or may not result in 6.647 representatives. Texas was accelerating through 2015, and slowing down since. The recovery of the price of oil may push the rate back up.

Annual estimated change: 404K, 434K, 401K, 475K, 501K, 450K, 400K (2011 through 2017).

When the 2018 county estimates are released next month, I will refine the 2020 projections.

Within counties, I used the 2013-2017 ACS block group estimates. Since these are based on a monthly sample over 5 years (60 monthly samples), they represent an average over the period. Assuming steady change, they can be treated as an estimate of the July 2015 population (midpoint of the 60 months).

I did not attempt to project these to 2020, since I thought that the census tracts and block groups had changed since 2010. In fact, the tracts and block groups are unchanged, even though there may be significant growth. One census tract in the 2013-2017 ACS has a population of 65,000.

I did scale the 2013-2017 estimates to the 2020 county projected populations. For example, the total population of block groups for Travis County (2013-2017) ACS was 1,176,584. This is in good accordance with the Census 2015 estimate of 1,178,292 (0.145% difference).

The 2013-2017 ACS populations were scaled so that they totaled the 2020 projected county population of 1,313,448, That is each block group population was scaled by 1313448/1176584 (1.116). That is each block group was projected to increase by 11.6% from 2015 to 2020.

In fully developed areas, the population typically declines. In a newly developed area, young couples (20s) purchase the homes, because they want a more modern home with more bedrooms, a yard, and access to better schools. At first the population increases, as children are added. As the children reach adulthood, they leave home either to college or the city.

The original couple tends to stay in the same house. Originally, they may have been economically stretched to afford the house. Later, they may seek stability for the children's education. Even if they change jobs, they likely considered the cost of commuting. As the children leave, the household population decreases, as does the neighborhood population.

The original couple is settled in by now, and will only leave when they retire, move to senior living, or die. It is only then that a new couple can move in. Unless there is multi-family added, the population will decline.

In older developed areas, the lowest quality housing stock may be abandoned. If your $200,000 house won't sell. your $180,000 house will. If you are moving to another city, you likely need the cash rather than hanging on for three years to sell. The $180,000 house becomes a $160,000 house, and so on, until you get to the $60,000 bungalow. It will either be abandoned, converted to rental, or torn down and replaced by a $300,000 4-story behemoth that fits on the lot barely.

Upgrading the bungalow is expensive. "Charming and quaint" translates to "small, wear long johns, and pile on quilts in winter, sweat in summer, and get used to charging your phone at Starbucks". There is a reason the bathroom is not mentioned in the advertisement.

So a block group that has declined by 5% between 2010 and 2015 will be projected to increase by 11.6% from 2015 to 2020, and a block group that increased by 23% between 2010 and 2015 will be projected to increase by 11.6% from 2015 to 2020.

This will have an effect on intracounty boundaries, but not intercounty apportionment. That is, based on the 2020 projections, there could still be a Blanco-Travis-Hays district, but the boundaries in Travis and Hays would be somewhat/slighty different.

I may try to project 2020 populations for block groups, now that I know that the 2010 base populations use the same areas.

The Census Bureau guidance for dividing or eliminating census tracts is to use the 2010 populations. That is census tracts that were too large or too small in 2010 have continued to be used. If a different population is used, the Census Bureau would require justification.

The Census Bureau expects census tracts to have a population between 1200 and 8000 persons with an ideal of 4000. Since the Census Bureau produces ACS data for formal political subdivisions, which in Maine includes towns. Small census tracts are not really a problem. If a town has a population of 500, it will have a greater sample rate in order to provide staistical reliability. The state liaison for census tracts is likely the same as for VTD's and census blocks.

[lfromnj]

Didnt Beto come pretty close in the core Lubbock district? it was something like a 10 point loss or something like that.

[Kevinstat]

...
HD-27 (+1.69%) Fort Bend(21%, 5% of county), Wharton(21%), Matagorda(18%), Calhoun(11%), Colorado(11%), Lavaca(10%). Cities Bay City, Port Lavaca, and El Campo.

You forgot Jackson (8%) (or possibly rounding to 7% of the district since the district is +1.69%).

[Kevinstat]

Didnt Beto come pretty close in the core Lubbock district? it was something like a 10 point loss or something like that.

According to Emil Shabanov (@emilshabanovTX, who made the map of the Cruz/O'Rourke Senate race by House district), the result in the current State House district entirely in Lubbock County (HD 84) was Cruz 55.90%, O'Rourke 43.27% (a 12.63% marin for Cruz).  That's a lot better for Democrats than Trump 59.58%, Clinton 35.12% in 2016, but it's still not particularly close.  It's possible Beto did a bit better in the cure Lubbock district that Jimrtex has drawn, but I imagine it would still have a pretty strong Republican lean (Cruz seemed to run well behind other Republicans).

[lfromnj]

Also it is now physically possible to create a D district in the panhandle in Texas. Obviously not allowed due to the county split rule and its a Blatant D gerrymander district

https://twitter.com/mattmxhn/status/1112211634305552384

[jimrtex]

...
HD-27 (+1.69%) Fort Bend(21%, 5% of county), Wharton(21%), Matagorda(18%), Calhoun(11%), Colorado(11%), Lavaca(10%). Cities Bay City, Port Lavaca, and El Campo.

You forgot Jackson (8%) (or possibly rounding to 7% of the district since the district is +1.69%).

Thanks. I calculated an ordered list of counties and pasted this into the list. My must have cut one too few, or otherwise, lost Jackson.

[Oryxslayer]

Its also possible to get a beto seat in Lubbuck...

https://twitter.com/emilshabanovTX/status/1112472921560285185

[Kevinstat]

Its also possible to get a beto seat in Lubbuck...

https://twitter.com/emilshabanovTX/status/1112472921560285185

For the benefit of those who don't want to click on the link for whatever reason, here's a map of that hypothetical Beto House district entirely in Lubbock County:



Interesting that the most Democratic all-Lubbock County House district discovered so far (that I'm aware of) extends well outside the City of Lubbock (that is itself too large for one House district) to a corner of the county.  Any particular factors that make southeastern Lubbock County (apart from Slaton or the portion thereof not included in that district) Democratic or at least less Republican than parts of Lubbock itself?

[dpmapper]

Interesting that the most Democratic all-Lubbock County House district discovered so far (that I'm aware of) extends well outside the City of Lubbock (that is itself too large for one House district) to a corner of the county.  Any particular factors that make southeastern Lubbock County (apart from Slaton or the portion thereof not included in that district) Democratic or at least less Republican than parts of Lubbock itself?

Probably has less to do with that part of Lubbock County in particular, and more to do with the remaining parts of Lubbock itself being ultra-Republican. 

[jimrtex]

Interesting that the most Democratic all-Lubbock County House district discovered so far (that I'm aware of) extends well outside the City of Lubbock (that is itself too large for one House district) to a corner of the county.  Any particular factors that make southeastern Lubbock County (apart from Slaton or the portion thereof not included in that district) Democratic or at least less Republican than parts of Lubbock itself?

Probably has less to do with that part of Lubbock County in particular, and more to do with the remaining parts of Lubbock itself being ultra-Republican. 

I wonder if that is even optimal.

The two Slaton precincts are 71% and 51%, which means that the Hispanic side of town is to the east. But to reach that they had to go through an 81% Cruz precinct. They had a choice of two precincts that were 81% Cruz and took the one with fewer votes.

It is possible that they were having to take 60% Cruz precincts. They were also using 2010 census figures. Democratic-leaning precincts are likely to have slower growth, plus Lubbock was not keeping up with the 18% statewide growth.

[jimrtex]

HD-88 is entirely within the city of Lubbock, with about 1/4 of the city in HD-87. The current district has a strange intrusion from the south that reaches the campus of Texas Tech, and an extrusion that reaches (the former) Reese AFB. These are unlikely to reflect partisan concerns. They are either to preserve separate districts for incumbent representatives, or to give both districts part of the economic engines of the area.

I simplified the boundaries, such that HD-87 is more peripheral areas of Lubbock, which may help the smaller counties be more competitive in determining representations.

The real beneficiaries might of that might be people in say, Shallowater or Wolfforth (still in Lubbock County but outside the City of Lubbock itself, and between the City of Lubbock and at least some (all in the case of Shallowater) of the other counties in the 2-district conglomerate).  Of course the current conglomerate may be different, but I'm talking about the benefit of smoothing out the boundaries of the main Lubbock district as opposed to making minimal changes to that district and putting the rest of Lubbock County in with the other counties in your proposed (possibly new) conglomerate.

The current districts in west Texas are somewhat odd because the goal was to avoid pairing incumbents. If that was not possible, there was an effort to draw districts that were "fair" contests between two incumbents. That is part of the reason for the odd districts in Dallas County.

The current incumbent in the outer Lubbock district lives in the extreme tip of the extension into Lubbock. But he was not the incumbent in 2011. The senator from the area resigned to become chancellor at Texas Tech, and the former representative for the outer Lubbock district became senator.

The current representative for the district is the chair of the Republican caucus. If the legislature draws the districts, they won't use my map.

Growth in the county is to the south and west, and into two districts outside Lubbock ISD. Wolfforth Frenship is the only 6A school in the county, but Frenship ISD includes a big chunk of western part of the city. Districts sometimes manipulate the size of the high schools for classification reasons. Three of the Lubbock hogh schools are just below the 5A cap, with the 4th much smaller. If they were not evenly balanced, one or two might be among the smallest 6A schools. It likely reduces travel costs.

Frenship with a single high school keeps adding buildings. It is pretty momentous to add a second high school, particularly if the current high school is centrally located.

Lubbock-Cooper is now a 5A school and includes the southern fringe of Lubbock. In the four largest counties, I used the school districts to draw boundaries, because the current distrcts are so awful.

Most other places I simply cleaned up the boundaries.

[jimrtex]

I see that according to the projected decimal "quotas" you've been using (perhaps different from what you expect to happen in some cases, like in Travis County, but what you're using all the same), Harris County is the one county with a projected quota that is within 5% of more than one integer.  In fact Harris's 24.719 could be (25, -1.12%) (which you use, understandably), (24, +3.00%) or even (26, -4.93%).  Unless, that is, there is some court precedent forbidding one of those last two options (particularly the last one that is more than a whole Representative different from Harris County's "ideal" (fractional) number of State Representatives.  I'd be curious to know what the legal precedent is there.  With Bexar County being projected close to a range which would be within 5% of two integers (the smallest such range, interestingly enough), the ability to either round up or round down could be relevant if it could potentially allow the deviations in single-county districts/conglomerates add up closer to 0 (right now, it's very close but Travis County getting within 5% of 7 and/or Montgomery County ending up more than 5% over 3 could upset that) and thus make the math less tight elsewhere.  Although I guess in that case giving Bexar County an 11th district would make things even worse, but you could think of it as giving Bexar only 10 districts if its quota was 10.494 (4.94% greater than 10; only 4.60% less than 11).  I'd be interested to know about the legal possibilities here.

As you've noted, the last range of decimal numbers that isn't within 5% of any integer is the range from 9.45 (5% greater than 9) to 9.50 (5% less than 10).  Once you get up to 10.45 (5% less than 11; only 4.5% greater than 10), you have at first small but then larger and larger ranges of decimal numbers within 5% of two integers until the last range that isn't within 5% of two integers is the range from 18.90 (5% greater than 18) to 19.00 (5% less than 20).  The lowest range of decimal numbers within 5% of three integers (or the lowest range of decimal numbers that's within 5% of an integer that is a whole number or more off from the number itself) is the range from 20.90 (5% less than 22) to 21.00 (5% greater than 20).  The range of numbers within 5% of 24, 25 and 26 (the range that you project Harris County to be in) is half a "quota" long (from 24.700 to 25.200).

The constitution is clear how apportionment is to be done. You divide a county's population by the quota and give it that many representatives. Then you form additional districts from smaller counties and/or surpluses. Historically, many more counties were entitled to their own representative, and there were relatively few entitled to extra. The largest had three.

In 1900, there were 27 counties with a single representative. These 27 counties were collectively entitled to 26.662 representatives an average deviation of -1.3%.

Their deviation ranged from +22.2% to -21.6%, with a standard deviation of 12.0%. It is somewhat high because there was a distribution gap around 1.00 quotas. Only 4 counties were within 5%, but 11 were between 5 and 10%.

Counties with larger deviation were apparently cases where there were not adjacent counties that could be used to balance out the population. It does not appear to be a simplistic, any county between 80% and 120% of a quota received a single representative. That is, if there was an opportunity to reduce the deviation for a county entitled to 1.15 or 0.85 counties it was used, but otherwise, the slightly larger deviation was acceptable.

There were 13 counties with two or three representatives, for a total of 29 representatives. Collectively they were entitled to 29.083 representatives or an average deviation of +0.3%.

With 2 representatives: Lamar, Galveston, Hunt, Fannin, Collin, Ellis, Navarro, Travis, Bell, and Hill.
With 3 representatives: Harris, Grayson, and Bexar.

The deviation range was -8.8% to 13.0%, with a standard deviation of 6.9%. This is not too surprising since the deviation is divided between two or three representatives. A county entitled to 2.2 representatives would have a deviation of 10%, while one with 1.2 would have a 20% deviation.

There were 27 multi-county districts with a single representative. Collectively. they were entitled to 27.251 representatives an average of the 0.9%.

I modified a pair of these.

HD-54 consisted of Jackson, Calhoun, Victoria, Refugio, and Bee, and had a deviation of +36.5%. HD-109 consisted of Live Oak, Karnes, and Goliad, and had a deviation of -16.6%.

These were the extreme deviations for multi-county districts. The two districts rap around each other, so Bee is surrounded on three sides by HD-109, and Goliad is surrounded on three sides by HD-54. HD-109 is anomalously numbered. suggesting that it was a last minute change. Moving Bee to HD-109 results in:

HD-54 consists of Jackson, Calhoun, Victoria, and Refugio, and has a deviation of +3.1%.
HD-109 consists of Live Oak, Karnes, Bee, and Goliad, and has a deviation of +16.8%.

The difference between the two is is reduced from 53.1% to 13.7%. The overall improvement is so great, that I suspect either a drafting error or a scriviner's error.

With these changes, the range in deviation was from 33.1% to -15.7%, with a standard deviation of 13.1%.

The multic-county districts were mostly along the Gulf Coast and West Texas, where there were clusters of counties with insufficient population of their own. The largest deviations tended to be among more isolated counties. The largest deviation was for Fort Bend and Waller, two counties too small for their own representative, and without ready alternatives for pairing.

These three types of districts: (1) Single county, single representative; (2) Single County, multi-representative, and (3)  Multi-county, single representative; provided 85 of 132 representatives, in 152 of 243 counties. These forms of legislative districts were normative for the era, particularly for a rural state with a vast territory. While they were not as equal as current standards, they could certainly be considered equitable.

Note: At this time. the constitution provided an upper limit of 150 representatives. It was only after several decades at 150 that the current specification of exactly 150 representatives was enacted.

There were 9 instances of single counties apportioned one or more representatives. In addition, they were placed in a floterial district with one or two smaller counties based on their surplus. The average surplus was 0.463.

There were 3 instances of multiple counties apportioned one or more representatives, along with a floterial district for their collective surpluses. This might be considered the classical example of a floterial district.

Fayette (1.583), Gonzales(1.251), Bastrop(1.163) = Total (3.996)
Milam (1.718), Robertson(1.363) = Total(3.081)
Tarrant (2.268), Denton(1.226). Wise(1.174), Cooke(1.191) = Total (5.860)

The numbering of the last district (108) indicates it might have been a late addition. By themselves, the individual counties did not have an atypical deviation, but it might have been realized that they were all shortchanged a bit. Montague (1.074) and Parker(1.118) could have been added to the group, but this would have placed almost 240,000 persons in the district. While theoretically representing the surplus of the six counties, the representative would have been elected from the entire 6-county area.

The floterial districts comply with the Texas Constitution, but their elections are problematic.

[jimrtex]

Didnt Beto come pretty close in the core Lubbock district? it was something like a 10 point loss or something like that.

According to Emil Shabanov (@emilshabanovTX, who made the map of the Cruz/O'Rourke Senate race by House district), the result in the current State House district entirely in Lubbock County (HD 84) was Cruz 55.90%, O'Rourke 43.27% (a 12.63% marin for Cruz).  That's a lot better for Democrats than Trump 59.58%, Clinton 35.12% in 2016, but it's still not particularly close.  It's possible Beto did a bit better in the cure Lubbock district that Jimrtex has drawn, but I imagine it would still have a pretty strong Republican lean (Cruz seemed to run well behind other Republicans).

I might have made it 45% O'Rourke. Adding the intrusion from the south, increased the Cruz majority, but the areas stripped off from the west and south removed a lot of Cruz voters. Using the 2015 ACS estimate, likely underpopulated the Lubbock city district. So perhaps 44% O'Rourke.

[jimrtex]

Districts 91-101 are in Tarrant County.



The map has been cleaned up using Paint to match school district boundaries. Population is based on block groups, which do not match school district boundaries, but should be within a few percent.

Rather than trying to adjust current districts and undo any gerrymandering, the districts are based on school districts (ISD). Tarrant is entitled to 10.907 representatives. With 11 districts, this is an average deviation of -0.8% which gives flexibility in varying the district populations to match district boundaries.

Note: My districts are based on block groups, which do not align with school district boundaries. I assigned each block group to an ISD, generally based on area, but also to give better contiguity and shape to the districts. In an actual map, I would match the actual ISD boundaries. So while the black lines (school district boundaries) don't match the house districts (color), imagine that they do.

When I had to split a school district, I tried to match city boundaries of smaller cities. This would be an impossible task for large cities such as Fort Worth, Dallas, San Antonio, and Houston.

School district boundaries in Texas are sometimes irregular, which may reflect consolidation patterns, as well as school bus routes. But they are much more regular than city limits, and fairly static. Generally, change is annexation of a whole district by another, rather than piecemeal changes. School districts are wary of losing students, even if an area would be better served by the other district. Their school planning may have sited schools based on anticipated development.

School districts represent a local community of interest. They provide more direct services to residents, and have more impact than streets or police or fire protection. Football stadiums provide a local gathering place on Friday night.

The following table represents the share of the 11 house districts in Tarrant County (e.g. Fort Worth ISD has 2.843/11ths of the county population vs. 2.819/150ths of the state population).

Fort Worth .......... 2.843
Arlington ........... 1.996
Keller .............. 0.973
Hurst-Euless-Bedford  0.871
Mansfield ........... 0.812
Birdville ........... 0.764
Eagle Mountain ...... 0.544
Crowley ............. 0.499
Grapevine ........... 0.458
Northwest ........... 0.234
White Settlement .... 0.186
Carroll ............. 0.174
Burleson ............ 0.126
Everman ............. 0.121
Azle ................ 0.109
Castleberry ......... 0.105
Kennedale ........... 0.101
Lake Worth .......... 0.084

Aledo and Godley ISD extend into Tarrant, but do not dominate any block grpups.

We can apportion out districts as follows:

Arlington 1.979 (two districts)
Keller 0.965 (one district)
Mansfield, Burleson, and Kennedale 1.039 (one district)
Eagle Mountain-Saginaw, Northwest(Justin), Castleberry, and Lake Worth 0.967 (one district)
Fort Worth and Everman 2.963 (three districts)

Hurst-Euless-Bedford(HEB), Birdville, Grapevine-Colleyville, and Carroll(South Lake) 2.268 (two districts plus a surplus),
Crowley, White Settlement, and Azle 0.794 (major portion of one district). This also includes the portions of Aledo and Godley ISD's in Tarrant. This area is not contiguous, but is intended for grouping population. Everman could be included, but demographically, it is a better fit with Fort Worth.

Collectively, these last two areas are entitled to 3.062 districts, but are not contiguous. We will use Fort Worth ISD to buffer the population between the two areas.

Haltom City and Richland Hills in Birdville ISD have the population to eliminate the surplus from the northeast area. In turn, the western part of FWISD is transferred to the western area.

Fort Worth ISD is divided into three districts. An eastern district along with Everman ISD is intended to be a predominately black district, a successor to HD-95. The northern district is intended to be a predominately Hispanic district. a successor to HD-90, and also includes the cities of Haltom City and Richland Hills in Birdville ISD. The near-western district is entirely in Fort Worth ISD.

HD-91 (-2.39%) Tarrant (100%, 9% of County). ISD: Fort Worth (100%, 35% of ISD); Cities: Fort Worth, Westover Hills, Westworth Village, and Edgecliff Village.

HD-92 (-1.96%) Tarrant (100%, 9% of County). ISD: Fort Worth (88%, 31% of ISD), Everman (12%). Cities: Fort Worth, Forest Hill, and Everman.

HD-93 (-2.25%) Tarrant (100%, 9% of County). ISD: Fort Worth (73%, 25% of ISD), Birdville (27%, 35% of ISD). Cities: Fort Worth, Haltom City, and Richland Hills.

Arlington ISD is divided into two districts, generally along Collins Street. This ensures that all of Grand Prairie is in one district. The city of Grand Prairie straddles the Dallas-Tarrant line. Grand Prairie ISD is entirely in Dallas County.

HD-94 (+0.25%) Tarrant (100%, 9% of County). ISD: Arlington (100%, 51% of ISD). Cities:  Arlington, Pantego, and Dalworthington Gardens.

HD-95 (-2.14%) Tarrant (100%, 9% of County). ISD: Arlington (100%, 49% of ISD). Cities: Arlington and Grand Prairie.

Keller ISD has population sufficient for its own district. Fort Worth has annexed a large share of the district.

HD-96 (-3.39%) Tarrant (100%, 9% of County). ISD: Keller (100%). Cities: Fort Worth, Keller, and Westlake.

Mansfield, Burleson, and Kennedale ISD's have population sufficient for a district. Burleson ISD is only the portion in Tarrant. The cities of Arlington, Grand Prairie, and Fort Worth have annexed south into the district.

HD-97 (+3.12%) Tarrant (100%, 9% of County). ISD: Mansfield 78%, Burleson 12%, Kennedale 10%. Cities: Mansfield, Arlington, Grand Prairie, Rendon CDP, Fort Worth. Burleson, and Kennedale.

Hurst-Euless-Bedford (HEB) 0.871 and Birdville ISD's each have population sufficient for a significant portion of a district, but there is no smaller ISD to complement them. Instead a district is created between them, and the collective surplus on the the west and east ends is added to other districts.

Both ISD's contain a number of cities, and the split attempts to match city boundaries, with Haltom City and Richland Hills excluded on the west, and Euless on the east.

HD-98 (-1.28%) Tarrant (100%, 9% of County). ISD: Birdville (50%, 65% of ISD), HEB (50%, 57% of ISD). Cities: North Richland Hills, Bedford, Hurst, and Watauga.

Eagle Mountain-Saginaw, Northwest, Castleberry, and Lake Worth ISD's have a population equivalent to a district. A small portion of Fort Worth ISD along Eagle Mountain Lake is included to get the district population closer to a quota, and avoid an odd panhandle on a district in Fort Worth ISD.

HD-99 (-0.69%) Tarrant (100%, 9% of County). ISD: Eagle Mountain-Saginaw (54%), Northwest (23%), Castleberry (10%), Lake Worth (8%). Fort Worth (3%, 1% of ISD). Cities: Fort Worth, Saginaw, River Oaks, Sansom Park, Lake Worth, Blue Mound, Pecan Acres CDP, and Haslet.

Crowley, White Settlement, and Azle ISD's have population equivalent to 3/4 of a quota. A portion of the western part of Fort Worth ISD is added to connect Crowley ISD with White Settlement and Azle ISD's. Azle ISD is physically separated from the rest of northern Tarrant by Eagle Mountain Lake, accounting for the Idaho-shaped district. Small portions of Aledo and Godley ISD's that extend into the county are also included into the district.

HD-100 (+2.50%) Tarrant (100%, 9% of County). ISD: Crowley (48%), Fort Worth (23%, 8% of ISD), White Settlement (10%). Azle (11%), Aldeo (0%), Godley (0%). Cities: Fort Worth, Benbrook, White Settlement, Crowley, Azle, Briar CDP, Pelican Bay, and Lakeside.

Grapevine-Colleyville and Carroll ISD's have about 3/5 of a quota. The Euless portion of HEB ISD is added to reach the quota.

HD-101 (+0.13%) Tarrant (100%, 9% of County). ISD: Grapevine-Colleyville (45%), HEB (37%, 43% of ISD), Carroll (17%). Cities: Euless, Grapevine, Southlake, and Colleyville,

[Kevinstat]

The July 1, 2018 population estimates came out yesterday.  I'm curious if the new numbers, when plugged into your model, result in any counties crossing an 0.95n or 1.05n (n being an integer) House quota boundary for 2020 for quotas less than an even 10, or an n+0.5 boundary for quotas greater than 10.  I guess I'd also be interested if you had Bexar County crossing the 10.45 (5% under 11) boundary, or the 10.476 (harmonic mean of 10 and 11) or 10.488 (geometric mean of 10 and 11) boundaries (and of course the 10.5 boundary, which is both the arithmetic mean of 10 and 11 and 5% over 10).

(An earlier post of yours makes it pretty clear that you can't apportion say, Harris County a number of House districts other than the number its quota rounds to.  With greater mathematical awareness since the Texas Constitution was written I'd argue some flexibility might be in order if Bexar County's quota ended up being between 10.476 and 10.5, or at least if it was between 10.488 and 5, but I'm not a lawyer.  Dallas and Tarrant counties aren't close to being within 5% of more than one integer in your projections and Harris County isn't close to being in an n+0.5 neighborhood.)

[jimrtex]

I have had an initial look at the 2018 county estimates.

Four counties had notable changes in their annual rate of change based on 2010-2018 projections vs. 2010-2017.

Aransas went from +1.3% to +0.3%. With essentially all the gain from 2010-2018 wiped out by Harvey. Some of this may be temporary as the estimate was for July 2018, 10 months after Harvey. Many people may have still been living in temporary locations while homes and businesses were rebuilt. Last summer, they were earnestly advertising that they were open for business, in an area heavily dependent on tourists.

Concho went from -5.5% to +0.5%, as a result of the closure of a privately-owned prison that was contracted by the federal government. The 2017 estimate reduced the county population by about 1/3 (1400). The 2018 restored the 2017 number and also for 2018. I found many news articles about the closure, but none about re-opening. In addition, to the 1400 inmates, the facility provided 270 jobs, though many may have been commuting in from San Angelo and other counties. The company is advertising for employees, which appear to be contingent on a new contract. The jobs appear to be specialists, rather than guards (e.g. dental hygienist).

Ector went from 1.8% to 2.0%, and Midland went from 2.6% to 2.8%, as the price of oil returned to profitable levels. The Permian Basin is the only area in the country where $35/bbl oil is profitable. Ector returned to an increase of 5K per year that it had achieved earlier in the decade, after two years of losses. Midland had its best year of the decade (7.2) after two years of small increases.


15 counties had a decrease in projected entitlement of 0.005. I think that in general this reflects the overall decrease in the rate of growth in cities, from the beginning of the decade (a trend that is nationwide). There may also be a Harvey effect.

27 counties had an increase in projected entitlement of 0.005 representatives or more. These seem to be concentrated in suburban areas, but this may be a residual of the lower growth rate in cities. As the urban share declines, the share of everything else increases.

Losers:

Big Cities
Bexar -21 (-.021)
Dallas -61
Denton -13
Harris -154
Travis -34

Border
El Paso -18
Cameron -7
Hidalgo -31
Webb -11

Harvey
Aransas -13
Fort Bend -24
Galveston -17
Jefferson -10
Nueces -6
Orange -10

Suburban San Antonio
Atascosa +6
Comal +17

Suburban Austin
Bastrop +6
Williamson +5

Suburban Houston
Brazoria +9
Liberty +11
Montgomery +23

Suburban/Exurban Dallas
Collin +59
Ellis +17
Fannin +5
Grayson +11
Henderson +7
Hood +9
Hunt +11
Johnson +13
Kaufman +20
Navarro +5
Parker +18
Rockwall +8
Wise +8

Prison
Concho +10

Permian Basin
Ector +16
Midland +24

I-35
Bell +21
McLennan +11

Other
Rusk +11
Taylor +7

[Kevinstat]

Okay, so Harris County's quota of 24.565 (+/- 0.001 if you were using rounded change figures, not +/- anything (well, +/- 0.0005 technically) if you were calculating the change in the rounded figures) would not be within 5% of 26 like 25.719 would (not that that matters at all from what I gather), but it's still closer to 25 than to 24 so it would still get 25 seats.  Montgomery County, at 3.170 quotas, will have to have a partial forth district, compared to an even three districts in your earlier plan and two whole disticts and most of a third currently based on the 2010 numbers.  Ellis County will be even more agonizingly close to the 0.950 mark, at 0.947.  I don't see any counties crossing an 0.950n or 1.050n mark in these revised projections besides Montgomery (and Harris).  Travis County will be further from the 6.650 mark, at 6.613 rather than 6.647.

Collin + Denton is now even closer to the ideal population for 10 districts (at 9.991 quotas rather than 9.945).  And Bell + McLennon will now (with these projections) be too big to be joined in a 3-district combo (that you hadn't done because of leftovers such a pairing would have left) anyway (with 3.174 quotas rather than 3.142).  I was mostly just looking at single counties and if their "status" changed, so there may be some changes to what could and could not happen with these new figres that I haven't noticed.

[jimrtex]

Okay, so Harris County's quota of 24.565 (+/- 0.001 if you were using rounded change figures, not +/- anything (well, +/- 0.0005 technically) if you were calculating the change in the rounded figures) would not be within 5% of 26 like 25.719 would (not that that matters at all from what I gather), but it's still closer to 25 than to 24 so it would still get 25 seats.  Montgomery County, at 3.170 quotas, will have to have a partial forth district, compared to an even three districts in your earlier plan and two whole disticts and most of a third currently based on the 2010 numbers.  Ellis County will be even more agonizingly close to the 0.950 mark, at 0.947.  I don't see any counties crossing an 0.950n or 1.050n mark in these revised projections besides Montgomery (and Harris).  Travis County will be further from the 6.650 mark, at 6.613 rather than 6.647.

Collin + Denton is now even closer to the ideal population for 10 districts (at 9.991 quotas rather than 9.945).  And Bell + McLennon will now (with these projections) be too big to be joined in a 3-district combo (that you hadn't done because of leftovers such a pairing would have left) anyway (with 3.174 quotas rather than 3.142).  I was mostly just looking at single counties and if their "status" changed, so there may be some changes to what could and could not happen with these new figures that I haven't noticed.

I had only noted that Bexar and Travis were further away from gaining a whole district, and   the drop in Harris.

I had not noticed that Bell and McLennan were possible as a pairing. Based on my rule that quasi-floterial districts should be minimized, they would have been paired. This assumes that it would be possible to draw districts in eastern and western Texas that were all within 5%. This is in no way a certain, since the population of the other counties to the east (Freestone, Limestone, Falls, Milam) is 0.432, and to the west (Coryell and Bosque) is 0.471. One side would have to absorb the extra among many districts, while the other side would gain a seat and have to spread a deficit around.

Looking at short term trends, Ellis is almost certain to reach 0.950. Based on annual estimates, Ellis went from 0.894 in 2010 to 0.892 in 2015. It has gained to 0.938 in 2018. The projection based on eight years is 0.947. Based on the last 3 years, it would be around 0.960.

If Ellis were its own district, Johnson would take Bosque and Somervell. This would reduce the western districts a tad, which could be absorbed given the smaller populations in that area. Hill would replace Bosque in the McLennan district. That would force Freestone out of the district. But it likely could be accommodated.
.585
You are right about Montgomery. A likely candidate would be to pair it with Brazos.

Harris will likely fall well below 24.5. Harris has almost 1/6 the state population, and must gain 72K per year just to stay even. It has fallen below that the last 3 years.

2010  24.410
2011  24.447
2012  24.517
2013  24.662
2014  24.792
2015  24.895
2016  24.855
2017  24.702
2018  24.556

Harris gained 0.585 in 5 years, but has lost 0.339 in the next three years. The projection of 24.565 based on 8 years is unrealistic.

If Harris has 24.300 in 2020, that would be an average deviation +1.2% with 24 districts vs. -2.8% with 25 districts. Either is easily doable within a single county.

If we limit ourselves to the 28 districts in eastern and western Texas, the change would be from the current -0.2%, vs. -3.6% for 29 districts. The deviation range would increase when shifting a district from Harris to these rural areas (from 2.6% to  4.8%). Moreover it is harder to consistently hit -3.6%, since only 1.4% below the mean we hit a limit.

It is possible more districts could be involved reducing the average deficit, making it more feasible to shift the district away from Harris.

For now, I'll keep Ellis and Harris as is, rather than trying to do a global projection over a shorter period of time.

I will adjust the Montgomery districts, and any multi-county districts that are out of range. I'll take a look at shifting counties to improve equality for the updated projections.

Incidentally, one of the virtues of weighted representation is that these shifts can easily be accommodated. If we were confident of the estimates, the weights could even be updated annually.

[jimrtex]

As an initial step, I updated the 2017 estimate based on the 2018-vintage release. Most of the statistics used by the Census Bureau are lagging. For example, 2017 tax returns were not available when the 2017 estimates were made (the Census Bureau uses tax returns and Social Security benefits to estimate internal migration).

For its estimates uses a component of change for its estimates. It knows that there a N persons, aged A, sex S, race R from the Census. Each year, some will die, and some will move. The others will stay put and age a year. There is pretty complete data on deaths and births. Migration must be inferred. It is particularly hard to track people who move away, unless you can locate them at their new residence.

The following were the largest changes from the 2017 estimate (vintage 2018) minus the 2017 estimate (vintage 2017). That is the largest changes (corrections) to the 2017 estimates.

Bexar -5 (-.005)
Brazos +3
Cameron -4
Collin +6
Concho +8
Dallas +16
Denton -7
Fort Bend +4
Galveston -5
Harris +43
Hidalgo -15
Lubbock -3
Montgomery +4
Tarrant +9
Webb -5
Williamson -10