2020 Redistricting in Pennsylvania

[palandio]

I'm currently doing a 51-seat map for Pennsylvania. With 51 seats things are not that bad for Democrats:

17 seats in SEPA, of which 16 are tilt D to safe D (by 2012/2016 PVI), one is tilt R

5 seats in Allegheny County, of which 3 are lean to safe D, while the two R-leaning ones are trending D

Scranton (D+6)
Allentown (D+4)
Erie (D+4)
Reading (D+3)
Harrisburg (D+3)
Bethlehem (tossup)
Easton, Stroudsburg (tossup)
Lancaster (R+1)
Wilkes-Barre (R+3)
State College (R+5)


19 likely to safe R

[palandio]

I'm currently doing a 51-seat map for Pennsylvania. With 51 seats things are not that bad for Democrats:

17 seats in SEPA, of which 16 are tilt D to safe D (by 2012/2016 PVI), one is tilt R

5 seats in Allegheny County, of which 3 are lean to safe D, while the two R-leaning ones are trending D

Scranton (D+6)
Allentown (D+4)
Erie (D+4)
Reading (D+3)
Harrisburg (D+3)
Bethlehem (tossup)
Easton, Stroudsburg (tossup)
Lancaster (R+1)
Wilkes-Barre (R+3)
State College (R+5)


19 likely to safe R

Isn't this basically a state senate map?

You're right. So I should better draw a State Senate map with 50 seats instead of 51 seats.

What this probably means is that with a ~50-seat map the Democrats' geographic disadvantage is smaller than with 17 or 18 seats. The current Pennsylvania State Senate has of course 28 Republicans and 21 Democrats (22 elected as Democrats) which would point to a R advantage.

How do the single areas compare?

SEPA has 14 Democrats which is close to the ceiling except for the 6th where Tomlinson seems to be a strong R incumbent in a D-leaning seat.

Allegheny has 4 Democrats, the 38th and the 43rd seem to be tossups, as expected (one D-held, one R-held).

Scranton is D.
Reading is D.
The Allentown/Bethlehem/Easton area has only one Democrat, due to both the Bethlehem/Easton seat being a D pack and Browne being a strong incumbent in the 16th.

Erie hasd Republican Laughlin elected in a D-leaning seat in 2016.
Harrisburg is drawn less favorably for the Democrats than is could be, Republican DiSanto gained this seat in 2016.
Lancaster is in traditionally R territory.
Wilkes-Barre elected Yudichak, but he left the Democrats.

Hence most of the Democrats' disadvantage is due to three Republicans in D-leaning seats and some is due to the map being drawn unfavorably for them (Harrisburg, Allentown).

[palandio]

[...]
How much priority do you place on reducing the # of county splits?

That's a difficult question. Generally I would say that reducing county splits is desirable and that a map that has many more splits than necessary is in most cases a bad map. On the other hand a good map is defined by more than just reducing splits: Representing CoIs, compactness, reasonable responsiveness to swings, a certain degree of partisan balance, etc. The optimal map may in many cases have slightly more than the mininally possible number of county splits.

[palandio]

[...]

This is what I came up with for the state senate: https://davesredistricting.org/join/f38b31b1-1fa9-4412-9221-6347e1778457

Basic assumptions were clean lines, keeping urban areas together and maximising the number of VRA seats in SE PA. Twenty-two districts have a D PVI in 2016, 28 had an R PVI. The tipping point seat is a little over R+4. That might have improved a little in 2020 given trends, but it'll still be at least a couple of points R.

Yes, that map makes a lot of sense in so many places.

But it can also be tweaked into the Democrats' direction in a few places. I did a rearrangement of districts in Bucks/Montgomery and Allegheny in the following way (first comes EastAnglianLefty's proposal, then my own):




The new Bucks Central seat is mostly D-trending suburbs, it actually voted for Clinton (and it makes a lot of sense as a CoI). The district in Northern Bucks and Northern Montgomery serves mostly as a Republican vote sink, a case of quite blatant gerrymandering and I'm not sure whether it is a wise thing to do. The two long-ish districts in Montogomery are aesthetically meh, but I have seen much worse creations. The Southern Bucks district should be clearly D-leaning, but who knows with a R candidate like Fitzgerald or Tomlinson.




A setup that is at the same time a clear tweak in the Democrat's favor and very aesthetically pleasing. The South-Eastern district which has less favorable trends for the Democrats is given a safety margin. All Allegheny districts voted for Clinton. The main problem that I see is that African American influence is diluted compared to EastAnglianLefty's proposal.

What does this mean overall?

Democrats have 13 seats with a PVI >5 (11 in SEPA, 1 in Pittsburgh, 1 in Allentown)

Then you have:
14. Allegheny SE D+5 (Clinton +8)
15. Scranton D+5 (Clinton +0)
16. Bucks South D+4 (Clinton +7)
17. Harrisburg D+4 (Clinton +10)
18. Erie D+3 (Clinton +1)
19. Reading D+2 (Clinton +4)
20. Chester NE D+2 (Clinton +13)
21. Allegheny NE D+2 (Clinton +7)
22. Bethlehem, Easton EVEN (Clinton +0)
23. Chester SW EVEN (Clinton +6)
24. Allegheny NW EVEN (Clinton +6)
25. Allegheny SW EVEN (Clinton +5)
26. Lancaster R+1 (Clinton +2)
27. Bucks Central R+2 (Clinton +3)

That is 27 Clinton districts (although she won three of them by less than one percentage point). Seems like a level playing field to me.

[palandio]

The final tweak would be in NEPA. Allentown moves from Likely D into competitive territory (not sure if that's a good idea from a D point of view, Browne [R] won his Allentown seat). This is compensated by three seats moving slightly towards the Democrats (Lackawanna, Luzerne, Bethlehem) and another competitive seat being created (Easton and Stroudsburg).



Going by 2012/2016 PVI this would yield the following competitive districts:

13. Scranton D+6 (Clinton +4)
14. Allegheny SE D+5 (Clinton +8)
15. Bucks South D+4 (Clinton +7)
16. Harrisburg D+4 (Clinton +10)
17. Allentown D+3 (Clinton +9)
18. Erie D+3 (Clinton +1)
19. Reading D+2 (Clinton +4)
20. Chester NE D+2 (Clinton +13)
21. Allegheny NE D+2 (Clinton +7)
22. Bethlehem D+1 (Clinton +2)
23. Chester SW EVEN (Clinton +6)
24. Allegheny NW EVEN (Clinton +6)
25. Allegheny SW EVEN (Clinton +5)
26. Lancaster R+1 (Clinton +2)
27. Easton and Stroudsburg R+2 (Trump +6)
28. Bucks Central R+2 (Clinton +3)
29. Wilkes-Barre R+4 (Trump +17)

The main takeaway for me is the following: With 50 seats it is easier to draw a nice-looking map that gives Democrats even chances than with 17 seats. The main reason seems to be that the SEPA and Allegheny areas can be kept homogeneously D, while enough compact competitive districts can be created outside SEPA and Allegheny that would otherwise drown in the R sea.

[palandio]

If you want two districts in the Pittsburgh area with a D PVI (2012/2016 numbers), this can be done without splitting the city itself. The green district is D+6.97, the orange one is D+0.24.
(I have not tried with Pittsburgh+Beaver county yet, maybe this will work even better?)

[palandio]

Continuity map that axes one R district:

2012/2016 PVI, 2016 Clinton share of 2-party vote

1: R+0.60, Clinton 51.01%
The MontCo part is extended both northwards and southwards to exactly preserve the district's political lean.

4: D +2.96, Clinton 54.93%
6: D +1.30, Clinton 54.28%
Wassermann slightly rotates the 4th, 5th and 6th which allows him to preserve the Montgomery/Delaware county border and gives the 4th a nicer shape and more continuity with its current version. My solution preserves the Delaware/Chester county border and gives the 6th a nicer shape and more continuity with its current version. I guess that you have to swallow one of these bitter pills or go for a completely new solution.

7: R +1.47, Clinton 48.04%
8: R +0.61, Clinton 45.54%
Here the continuity solution would have been to expand the 7th's share of Monroe county at the expense of the 8th and to expand the 8th to Carbon county plus some additions in Luzerne county. My solution includes the nice 4-county combo for the 8th that avoids one county split. It is slightly more favorable for Cartwright and slightly less for Wild than the pure continuity solution.

9: R +6.10, Clinton 44.37%
More or less preserves the old D longshot district. The Dauphin/Cumberland/York/Lancaster 4-county combo is quite nice and fits perfectly with a continuity map. Lebanon county is a bit of a problem if you also put Berks with SEPA but I think that my solution for Lebanon county is still acceptable.

16: R +3.55, Clinton 47.83%
I tried to keep the changes in Allegheny county to an absolutely necessary minimum and was happy that I could do this without forcing an unnecessary additional county split. Lamb's district of course moves slightly to the right, as has to be expected from a continuity map, but not too much.

[palandio]

I was skeptical about Luzerne+Lackawanna+Monroe+Pike being exactly one district in 2020, but according to 2019 estimates they're really only ca. 100 inhabitants away from the perfect quota, so it's quite likely that this will be similar in 2020.

[palandio]

[...]

On that note, is there any reasonable way Lamb could be given a reasonably winnable seat without a pro-inc / Dem gerrymander? Most variations I see have Lamb's seat at around Trump +5 in 2020. Not that he couldn't win that, but he barely overperformed Biden this year and it seems like he is no longer above the partisan lean of his district a la his 2018 special election.

Depending on whether you consider the 2018 court map a Dem gerrymander or not in the Pittsburgh area, it is very easy to give Lamb a reasonably winnable seat:
- The line between Lamb's seat and the Pittsburgh seat in Allegheny county stays exactly as it is.
- The Pittsburgh seat takes in parts of e.g. Westmoreland county from the current 14th.
- The current 14th will include Washington, Green, Fayette, Somerset and most of Westmoreland. In fact the only county split of the successor of the 14th would be in Westmoreland with the Pittsburgh seat, hence avoiding an unnecessary county split.
- Lamb's seat could be extended into Butler or Lawrence.

An extension into Butler would move Lamb's seat from Trump +2.53 to Trump +4.26, going by 2016 results. An extension into Lawrence could move Lamb's seat to as low as Trump +3.26. I don't have 2020 numbers, but given that Lamb won by 2.2, his margin would go to ca. 0.5-1.5 percentage points.

[palandio]

[...]

On that note, is there any reasonable way Lamb could be given a reasonably winnable seat without a pro-inc / Dem gerrymander? Most variations I see have Lamb's seat at around Trump +5 in 2020. Not that he couldn't win that, but he barely overperformed Biden this year and it seems like he is no longer above the partisan lean of his district a la his 2018 special election.

Depending on whether you consider the 2018 court map a Dem gerrymander or not in the Pittsburgh area, it is very easy to give Lamb a reasonably winnable seat:
- The line between Lamb's seat and the Pittsburgh seat in Allegheny county stays exactly as it is.
- The Pittsburgh seat takes in parts of e.g. Westmoreland county from the current 14th.
- The current 14th will include Washington, Green, Fayette, Somerset and most of Westmoreland. In fact the only county split of the successor of the 14th would be in Westmoreland with the Pittsburgh seat, hence avoiding an unnecessary county split.
- Lamb's seat could be extended into Butler or Lawrence.

An extension into Butler would move Lamb's seat from Trump +2.53 to Trump +4.26, going by 2016 results. An extension into Lawrence could move Lamb's seat to as low as Trump +3.26. I don't have 2020 numbers, but given that Lamb won by 2.2, his margin would go to ca. 0.5-1.5 percentage points.

I see. I don't consider the current map a Dem gerrymander, but of all the fair maps possible, it is definitely a more Dem-favorable configuration.

My assumption was that the Pittsburgh seat would stay entirely in Allegheny, and in the process it would take in most of the Dem-friendly inner suburbs, thereby hurting Lamb. If it does expand into Westmoreland instead, then it would leave room for Lamb to take in the friendly suburbs that have supplied him his past victories.

https://davesredistricting.org/join/7451cc49-5088-4dea-95f2-d89dbea3c030

This was my take. I started with Pittsburgh and then tried to take in the rest of Allegheny whilst making it compact. If compactness and not trying to split Westmoreland is a priority, then I don't really see Lamb getting a favorable seat.

Yes, no doubt that under compactness considerations* and non-splitting considerations your arrangement or something very similar makes the most sense.

If you wanted to draw a Lamb seat under 2016 numbers that is as Dem-friendly as possible and still leaves a Pittsburgh seat entirely contained in Allegheny, you would put the South and East of Allegheny into Lamb's seat and the North and West into the Pittsburgh seat. Adding Washington and parts of Westmoreland would even give you a Clinton seat, although with less favorable trends than Lamb's current seat. The ugliness of this arrangement would mostly come from ripping the Washington-Greene-Fayette-Westmoreland area apart.

* My personal definition of compactness for districts differs slightly from the most common ones: Voters with a small geographic distance between them should be in the same district. E.g. for every two voters that are in the same district, the map gets a score that is proportional to e.g. minus the logarithmized geographic distance of these voters. Bonus score if the voters live in the same county and the same city. Neat borders and good shapes are the automatic products of this criterion, but they are not the criterion itself.

[palandio]

[...]

On that note, is there any reasonable way Lamb could be given a reasonably winnable seat without a pro-inc / Dem gerrymander? Most variations I see have Lamb's seat at around Trump +5 in 2020. Not that he couldn't win that, but he barely overperformed Biden this year and it seems like he is no longer above the partisan lean of his district a la his 2018 special election.

Depending on whether you consider the 2018 court map a Dem gerrymander or not in the Pittsburgh area, it is very easy to give Lamb a reasonably winnable seat:
- The line between Lamb's seat and the Pittsburgh seat in Allegheny county stays exactly as it is.
- The Pittsburgh seat takes in parts of e.g. Westmoreland county from the current 14th.
- The current 14th will include Washington, Green, Fayette, Somerset and most of Westmoreland. In fact the only county split of the successor of the 14th would be in Westmoreland with the Pittsburgh seat, hence avoiding an unnecessary county split.
- Lamb's seat could be extended into Butler or Lawrence.

An extension into Butler would move Lamb's seat from Trump +2.53 to Trump +4.26, going by 2016 results. An extension into Lawrence could move Lamb's seat to as low as Trump +3.26. I don't have 2020 numbers, but given that Lamb won by 2.2, his margin would go to ca. 0.5-1.5 percentage points.

I see. I don't consider the current map a Dem gerrymander, but of all the fair maps possible, it is definitely a more Dem-favorable configuration.

My assumption was that the Pittsburgh seat would stay entirely in Allegheny, and in the process it would take in most of the Dem-friendly inner suburbs, thereby hurting Lamb. If it does expand into Westmoreland instead, then it would leave room for Lamb to take in the friendly suburbs that have supplied him his past victories.

https://davesredistricting.org/join/7451cc49-5088-4dea-95f2-d89dbea3c030

This was my take. I started with Pittsburgh and then tried to take in the rest of Allegheny whilst making it compact. If compactness and not trying to split Westmoreland is a priority, then I don't really see Lamb getting a favorable seat.

Yes, no doubt that under compactness considerations* and non-splitting considerations your arrangement or something very similar makes the most sense.

If you wanted to draw a Lamb seat under 2016 numbers that is as Dem-friendly as possible and still leaves a Pittsburgh seat entirely contained in Allegheny, you would put the South and East of Allegheny into Lamb's seat and the North and West into the Pittsburgh seat. Adding Washington and parts of Westmoreland would even give you a Clinton seat, although with less favorable trends than Lamb's current seat. The ugliness of this arrangement would mostly come from ripping the Washington-Greene-Fayette-Westmoreland area apart.

* My personal definition of compactness for districts differs slightly from the most common ones: Voters with a small geographic distance between them should be in the same district. E.g. for every two voters that are in the same district, the map gets a score that is proportional to e.g. minus the logarithmized geographic distance of these voters. Bonus score if the voters live in the same county and the same city. Neat borders and good shapes are the automatic products of this criterion, but they are not the criterion itself.

So, say you draw a district covering Somerset, Washington, Greene, Fayette, and as much of Westmoreland as possible, this would help Lamb?

If the rest of Westmoreland is used to fill up the population deficit of the Pittsburgh district, then yes, of course would this help Lamb. Because otherwise, as kwabbit said, the Pittsburgh district would need to take in some of the most Dem-friendly areas from Lamb's seat.

[palandio]

https://davesredistricting.org/join/bee23a27-3796-4f20-99da-98550eebd85a

Well, every map has its advantages and disadvantages, but what you are doing to the Allentown area deserves a big penalty under every immaginable metric. It might help you with the rest of the map and I give you credit for exploring the idea, but I think that the price is too high.

[palandio]

Two questions:

1) Why do folks seem to be assuming that Lamb's district won't be made more Democratic by giving it some more Democratic territory in Allegheny County while adding some more Republican turf (ex: taking parts of northern Alleghany County and/or some of southern Butler County) to Doyle's district?  I mean, yes, that'd be something you'd see in a Democratic map, but it's not so unreasonable that it'd be likely to give pause Democratic-leaning court.  I'd argue this is more likely than Lamb's district gaining southern Butler County (assuming it doesn't cause too many issues elsewhere in the map).  When in doubt, the PA SC's map is gonna favor the Democrats and this seems like a perfect example of somewhere that could happen.

2) Does anyone know why the "don't split Bucks County" rule has persisted for so long?  I'm not saying I expect it to change; I'm just curious.

Wouldn't it be more efficient to give Lamb all the Dem turf directly east of Pittsburgh, remove Butler, and turn his district into a Beaver+all of Allegheny as needed, and then give Doyle parts of Washington County instead?

If you go for a "least change" map in Allegheny, you can essentially view Lamb's, Doyle's and Reschenthaler's districts as a chain. All of these districts need to expand, but you can put the options into an order (by geographic similarity and by Dem-friendliness, not necessarily by quality):
a. Like you say, preserve the northern border of Lamb's district (or even remove the small parts of Butler that it contains), give him more areas around Pittsburgh (he can't get all areas to the east of Pittsburgh because otherwise he would have to give up on Mt. Lebanon). Doyle gets areas in Washington or Westmoreland County. Reschenthaler's district would then gain Somerset and Bedford County. This would move Lamb's district ca. 4-5 points to the left (ca. Clinton +2 in 2016).
b. Like I proposed earlier in the thread, roughly preserve the border between Lamb's and Doyle's district. Doyle's district would need to slightly expand into Washington or Westmoreland County. Reschenthaler's district would gain Somerset County. Lamb's district would expand into Butler or Lawrence County. This would move Lamb's district ca. 0.7-2 points to the right.
c. Preserve the border between Doyle's and Reschenthaler's district. Expand Lamb's district to include either Lawrence and minor parts of Butler or most parts of Butler. This would move Lamb's district ca. 10 points to the right (if you want to keep Mt. Lebanon in his district).
c2. Expand Reschenthaler's district to include all of Westmoreland, but also parts of Allegheny, but no parts of Somerset. Not that different from c.

All variants would avoid a county split at some border (a. either at the northern border of Lamb's district or by Reschenthaler's district including all of Somerset and Bedford, b. by Reschenthaler's district including all of Somerset, c. at the border between and Doyle's and Reschenthaler's districts, c1. at the eastern border of Reschenthaler's district).

Going by compactness metrics alone I get why many are proposing c. or c1. But I think that particularly b., but maybe even a. are completely defensible from a "continuity" or "partisan fairness" standpoint.

[palandio]

Northern Allegheny is about as GOP as Butler and makes that fairly difficult.

Maybe a few precincts in Cranberry Township, Butler, are comparable to Northern Allegheny, but overall no. Northern Allegheny has become much more Dem-friendly than Butler.

Two questions:

1) Why do folks seem to be assuming that Lamb's district won't be made more Democratic by giving it some more Democratic territory in Allegheny County while adding some more Republican turf (ex: taking parts of northern Alleghany County and/or some of southern Butler County) to Doyle's district?  I mean, yes, that'd be something you'd see in a Democratic map, but it's not so unreasonable that it'd be likely to give pause Democratic-leaning court.  I'd argue this is more likely than Lamb's district gaining southern Butler County (assuming it doesn't cause too many issues elsewhere in the map).  When in doubt, the PA SC's map is gonna favor the Democrats and this seems like a perfect example of somewhere that could happen.

2) Does anyone know why the "don't split Bucks County" rule has persisted for so long?  I'm not saying I expect it to change; I'm just curious.

Wouldn't it be more efficient to give Lamb all the Dem turf directly east of Pittsburgh, remove Butler, and turn his district into a Beaver+all of Allegheny as needed, and then give Doyle parts of Washington County instead?

That’d also work and just underscores my point about how easy it is to shore up Lamb.

https://davesredistricting.org/join/f3410140-478f-48b0-ba35-e8acd88c5f2b
thoughts on this map?

Putting Cambria into Reschenthaler's district avoids one more county split at the cost of compactness. Personally in this case I'd prefer compactness, but that's debatable.

Lamb would probably prefer Mount Lebanon to stay in his district, but as we agreed in another thread, incumbents' wishes should not take priority over everything else.

[palandio]

Exactly. If you want to go for clean maps, then there are essentially two possibilities:

1) Wild gets Stroudsburg. Then the 2020 numbers are:
Wild: ca. Biden +5
Cartwright: Trump +7.5 (up to Trump +10 with more R-friendly lines)

2) Cartwright gets Stroudsburg. Then the 2020 number are:
Wild: ca. Biden +1
Cartwright: Trump +3 (up to Trump +5 with more R-friendly lines)

Somewhere in the range of Biden +5 to Biden +6 is the reasonable maximum for Wild. Any attempt to shift more democrats from Cartwright to Wild (making Cartwright's seat Trump +12 or so) would involve very ugly gerrymandering.