Is there a systematic bias in Muon2's erosity rules for macro chopped counties?

[Torie]

The rules seems to reward macro-chops that run along county lines, because there is no erosity penalty if a municipality in the macro-chopped county is in a different CD than the municipality adjoining it in a different county.

So the top map has a worse erosity score than the bottom map (assume that the bottom county (Montco), has towns that are all squares and all have equal populations.



And the real world result seems to be that the map that has a macrochop into Montco running along the Berks County line all the way to Chester County, has a better erosity score than the map where the macro-chop hugs the Bucks County line.   Such a result seems facially unfortunate, particularly where everything across a county line that the macro runs along is in a different CD (which is the case with the top PA map where the nexus between the macro-chop and Bucks is so attenuated, and about 90% of the lines of a chop lie next to a foreign CD).

[jimrtex]

The rules seems to reward macro-chops that run along county lines, because there is no erosity penalty if a municipality in the macro-chopped county is in a different CD than the municipality adjoining it in a different county.

So the top map has a worse erosity score than the bottom map (assume that the bottom county (Montco), has towns that are all squares and all have equal populations.



And the real world result seems to be that the map that has a macrochop into Montco running along the Berks County line all the way to Chester County, has a better erosity score than the map where the macro-chop hugs the Bucks County line.   Such a result seems facially unfortunate, particularly where everything across a county line that the macro runs along is in a different CD (which is the case with the top PA map where the nexus between the macro-chop and Bucks is so attenuated, and about 90% of the lines of a chop lie next to a foreign CD).

How are the boundaries between the chops and Lehigh, Berks, and Chester scored?

Generally, we have been ignoring the external boundaries of the state, since these are common to all plans, and for Muon's method would be difficult to score, since we might also have to take into account counties in adjacent states, and the highways that cross the state line. It doesn't really matter for a total measure. It might have some effect on a measurement of an individual district (such as the pistol district in Iowa), but we might not care if a odd-shaped district permits the other districts to be more regular.

But it sounds like you are saying that it is only the inner Montco boundary that counts for the chop?

Perhaps each district could be scored independently. The statewide erosity would be the sum of the district erosity divided by 2. You could then have heightened scrutiny within chops.

So for District 6, you would have a count of 1 or 2 from Berks into the two Montco districts, and 1 or 2 from Chester into the two Montco districts. But for district 3, you would count all the town boundary crossings into Berks and Chester, as well as those into Montco District 4. The crossing of the Bucks-Montco line within District 3 does not count.

[Torie]

Yes, it seems to me in a macro chopped county,  that if there is pavement connecting the two subdivisions, then a county line (as opposed to a state line) dividing them should not cleanse a chop if across the county line is a foreign CD. There would also be an incentive to chop Chester vertically rather than the facially superior horizontal chop, in order to hug county lines better, and get more chop cleansing. There is also no incentive to append the chop into Montco with the one into Philly for the same reason.  

[jimrtex]

Yes, it seems to me in a macro chopped county,  that if there is pavement connecting the two subdivisions, then a county line (as opposed to a state line) dividing them should not cleanse a chop if across the county line is a foreign CD. There would also be an incentive to chop Chester vertically rather than the facially superior horizontal chop, in order to hug county lines better, and get more chop cleansing. There is also no incentive to append the chop into Montco with the one into Philly for the same reason.  

Of course I would not classify chops and treat them differently, but would only be concerned about how well the statewide plan provides practicable equality among districts. The split of Montco could be handled independently.

[Torie]

Here is the map for the whole state as I would like to draw it.

[jimrtex]

Here is the map for the whole state as I would like to draw it.

17 districts, 2020 projected populations?

[Torie]

Correct.

[jimrtex]

Corrected to use 2016 estimates

These are the population estimates for 2020/17 representatives. Each county was independently projected based on the 6-1/4 year growth from April 2010 to July 2016, forward to 2020.

UCCs:

Philadelphia (5.493 quotas). Will need about 1/2 a quota added or chopped.

Pittsburgh (3.006). Almost perfect for 3 districts, except it cuts off Greene.

Allentown (0.896).

Lancaster (0.728).

Harrisburg   (0.707).

Scranton (0.693)

York (0.595)

Ideally each of these five will form the core of their own district.

Reading (0.552)
Erie (0.362)
East Stroudsburg (0.217)
State College (0.220)
Chambersburg (0.207)
Johnstown (0.173)
Lebanon (0.188)
Youngstown, OH/Mercer (0.147)
Williamsport (0.152)
Gettysburg (0.136)
Bloomsburg (0.112)

These will be treated as separate counties, other than Bloomsburg UCC, which consists of two counties: Columbia and Montour, that will collectively be treated as a single county.

[jimrtex]

Updated: Projections based on 2016 estimates.

Here is a first cut, at county-based districts.



Berks is added to the Philadelphia UCC to produce a 6-district region. The division into districts is the best possible prior to division of counties. Philadelphia has two districts, while the next three counties, Montgomery, Bucks, and Delaware have one district each. The final district is assigned to Chester and Berks.

Greene gets added to the Pittsburgh UCC because it is cutoff. Placing Westmoreland with Allegheny in the two-district subregion would decrease the deviation a bit, but would make the other district less compact as it would wrap around from Butler to Fayette.

Carbon is the best fit for the Allentown UCC.

Lancaster, York, and Harrisburg UCC's are expanded outward to avoid conflict. The York district takes in Hunterdon rather than Bedford, since this provides a better community of interest for PA-16.

The Lancaster district is short, but the combined Philadelphia-Berks region has a small surplus, that can be used to add to the Lancaster district.

This version goes east from the Scranton district.

6 districts in Philadelphia, 3 in Pittsburgh, and 1 each for Lancaster, York, Harrisburg, Allentown, and Scranton total 14 districts, leaving 3 districts for the remainder of the state, which pretty much draw themselves. It appears that Centre has to be placed in the southwestern PA-16 district.

Shifting parts of Philadelphia and Montgomery into Bucks would result in four districts with an average deviation of 1.1%. Philadelphia gets divided north/south.

Shifting a small part of Allegheny to PA-9 produces a 1.8% deviation for three districts. Downstream and northern parts of Allegheny will be placed with Butler and Beaver, with the area placed in PA-9 likely being in the Mon Valley.

Making adjustments among PA-5, PA-6, and PA-13 produces a -2.8% deviation for the three districts. The most suburban areas of Chester will be moved to PA-5, and areas in southwestern Chester added to the Lancaster district. The remainder of Chester, generally to the north and west will be placed in a district in which Berks will form a small majority.

Assuming ideal splits, the standard deviation of the deviations is a mere 1.87%, and only four counties are divided, including the three largest which each have more than enough population for one district and must be split, and one additional suburban county.

[jimrtex]

Modified to use 2016 estimates for projections

This alternative shifts the Scranton district towards the west, which may be a little better community of interest.



The shift of Montour, Columbia, and Schuylkill, leaves Northumberland somewhat isolated. So Northumberland is moved to the Harrisburg district, requiring some further swapping of other counties.

Overall, this plan reduces the standard deviation of deviation to 1.73%.

Swapping Clarion for Elk and Cameron (not shown) would change the deviation of PA-15 to -1.1%, and of PA-16 to -0.5%, and the standard deviation of deviations to 1.71%. I think this is a negligible improvement.

[Torie]

Why didn't you use the population estimates for July 1, 2016?

[muon2]

The rules seems to reward macro-chops that run along county lines, because there is no erosity penalty if a municipality in the macro-chopped county is in a different CD than the municipality adjoining it in a different county.

So the top map has a worse erosity score than the bottom map (assume that the bottom county (Montco), has towns that are all squares and all have equal populations.

It depends on the connections across the county line. That is, how many of the south county subunits have regional connections to the north county. If they all did (and all adjacent subunits are locally connected), then the top plan would have lower erosity (13 to 14). It would be clearer if the subunits for the north county were not shown in the maps, since there is no chop there.

I think the issue arises when one focuses on the northern CD, which is arguably more compact in the top plan. However, the southern CD is more compact in the bottom plan. The erosity is balancing both shapes and since there's a macrochop and a county line the result depends on the number of connections across that line.

[jimrtex]

Why didn't you use the population estimates for July 1, 2016?

I was thinking that only the state estimates were out for 2016, since there were 51 data tables for estimates, and I conflated that with the number of states plus DC.

Fixed for the first map. Other maps will be forthcoming.

[jimrtex]

The rules seems to reward macro-chops that run along county lines, because there is no erosity penalty if a municipality in the macro-chopped county is in a different CD than the municipality adjoining it in a different county.

So the top map has a worse erosity score than the bottom map (assume that the bottom county (Montco), has towns that are all squares and all have equal populations.

It depends on the connections across the county line. That is, how many of the south county subunits have regional connections to the north county. If they all did (and all adjacent subunits are locally connected), then the top plan would have lower erosity (13 to 14). It would be clearer if the subunits for the north county were not shown in the maps, since there is no chop there.

I think the issue arises when one focuses on the northern CD, which is arguably more compact in the top plan. However, the southern CD is more compact in the bottom plan. The erosity is balancing both shapes and since there's a macrochop and a county line the result depends on the number of connections across that line.

I'm confused how this is to be scored.

Since the southern county is Montco, add in Berks across its western end, Lehigh and Northampton west of Bucks; Chester  and Delaware south of Montco, and Philadelphia to the east of Montco.

[Torie]

The rules seems to reward macro-chops that run along county lines, because there is no erosity penalty if a municipality in the macro-chopped county is in a different CD than the municipality adjoining it in a different county.

So the top map has a worse erosity score than the bottom map (assume that the bottom county (Montco), has towns that are all squares and all have equal populations.

It depends on the connections across the county line. That is, how many of the south county subunits have regional connections to the north county. If they all did (and all adjacent subunits are locally connected), then the top plan would have lower erosity (13 to 14). It would be clearer if the subunits for the north county were not shown in the maps, since there is no chop there.

I think the issue arises when one focuses on the northern CD, which is arguably more compact in the top plan. However, the southern CD is more compact in the bottom plan. The erosity is balancing both shapes and since there's a macrochop and a county line the result depends on the number of connections across that line.

I'm confused how this is to be scored.

Since the southern county is Montco, add in Berks across its western end, Lehigh and Northampton west of Bucks; Chester  and Delaware south of Montco, and Philadelphia to the east of Montco.

As I understand it, when a county is macro-chopped, you zoom in, and the subdivisions become quasi counties themselves, except that erosity road cuts internal to the macro-chopped county need only pavement roads, while road cuts across county lines, require state highways to get erosity penalty points, just like road cuts between counties not macro chopped, require a state highway to be cut that runs between the county seats to get a penalty point. So it pays to minimize the road cuts that are internal to the macro chopped county, which means since any pavement will do, minimizing the subdivisions that touch each other that are in different CD's within the macro-chopped county is the way to get a better score.

[jimrtex]

This is a third variant. It brings the Scranton district further down the Susquehanna, in exchange for Schuylkill being added to the Harrisburg district.



The standard deviation of deviations is 1.76%. A potential swap is Jefferson for Warren.

[jimrtex]

Corrected Map

This is a fourth look, moving Centre into PA-17. It has a standard deviation of deviations of 1.79%

[jimrtex]

This is the division of Allegheny County.



PA-8 takes the portion of the county north of the Allegheny River (except the portion of Pittsburgh, former Allegheny, that lies north of the river; and areas downstream along the Ohio River on both the north and south side of the river. These areas fit reasonably well on a community of interest basis with Butler and Beaver counties.

PA-9 takes a small area along the Monongahela River as it enters the county, generally east of the river. It jumps ahead a bit to take McKeesport based on population balance.

The three districts are not quite equal, changing the overall standard deviation (for this plan) from 1.786% to 1.790%.

Zoomed out to show all three districts.

[jimrtex]

This is the division of Chester:

To PA-5: 167K along the Delaware line and closer to Philadelphia
To PA-6: 320K bulk of county with adjacency to Berks (this is 3/7 of district)
To PA-13: 41K in southwestern corner adjacent to Lancaster



Zoomed out to whole districts:

[jimrtex]

This is the provisional division of Montgomery.



Because it is the intent that the surplus above two seats in Philadephia be transferred to Bucks, it makes sense to also move an area in Montco near the Bucks-Montco-Philadelphia intersection.

[jimrtex]

This is the division of Philadephia.


The 2015 5-year ACS was used to estimate census tract populations. This was assumed to represent the population as of July 2013 (midpoint of 5-year period). The aggregate population of the three district areas was calculated from 2010 Census and 2015 ACS, and then an annual growth rate for 3.25 years was used to project a population for 2020. Finally, the 2020 projections were controlled to the 2020 county estimate.

For the three district areas, the projected change are:

PA-1 3.46%
PA-2 6.36%
PA-4 -7.83%
Philadelphia: 4.33%

The difference in growth rate between PA-1 and PA-2, the two districts contained in Philadelphia, results in PA-2 adding about 17,000 more people over the decade. The decline in PA-4 is relatively negligible since the area is only about 1/10 of the population of either PA-1 or PA-2.

This is the 5-county Philadelphia Metro Area


PA-1 and PA-2 have deviations of +1.1% and +1.5%, respectively. Statewide standard deviation is 1.796% (up from 1.786% due to non-perfect division of counties).

Statewide map showing county splits:


Districts PA-1 to PA-10, and PA-13 are unchanged in all maps). Four arrangements of PA-11, PA-12, and PA-14 to PA-17 are illustrated up thread.