Based on 2012 results only, which state gerrymander flipped the most seats?

[muon2]

To jump back to Torie's first IN offering, I took it in the direction of reducing erosity first and chops second. All UCCs are maintained with minimal chops. Marion has two chops and a township microcrop, plus there is one other chop to equalize population. That's a chop score of 4 including the microchop.  The erosity score is 69 (compared to 73 for Torie's later offering with chop of 3).

The partisan breakdown is 1D, 1d, 1e, 3r, 3R. That make 5 competitive districts out of 9. I don't know if the Dem's want a plan where they could get as many as 6 CDs (7 with an R+6 CD-8), but could get shut out to one CD if the D+3 CD-7 swings against them. It's hard to judge flips with so much at play.

[Torie]

To jump back to Torie's first IN offering, I took it in the direction of reducing erosity first and chops second. All UCCs are maintained with minimal chops. Marion has two chops and a township microcrop, plus there is one other chop to equalize population. That's a chop score of 4 including the microchop.  The erosity score is 69 (compared to 73 for Torie's later offering with chop of 3).

The partisan breakdown is 1D, 1d, 1e, 3r, 3R. That make 5 competitive districts out of 9. I don't know if the Dem's want a plan where they could get as many as 6 CDs (7 with an R+6 CD-8), but could get shut out to one CD if the D+3 CD-7 swings against them. It's hard to judge flips with so much at play.

Other than CD-1, CD-7 was probably the one CD in Indiana that did not trend hard to the Pubs in 2012, and no doubt over time it will continue to trend Dem, as the Pubs thin out in Marion County.

[muon2]

To jump back to Torie's first IN offering, I took it in the direction of reducing erosity first and chops second. All UCCs are maintained with minimal chops. Marion has two chops and a township microcrop, plus there is one other chop to equalize population. That's a chop score of 4 including the microchop.  The erosity score is 69 (compared to 73 for Torie's later offering with chop of 3).

The partisan breakdown is 1D, 1d, 1e, 3r, 3R. That make 5 competitive districts out of 9. I don't know if the Dem's want a plan where they could get as many as 6 CDs (7 with an R+6 CD-8), but could get shut out to one CD if the D+3 CD-7 swings against them. It's hard to judge flips with so much at play.

Other than CD-1, CD-7 was probably the one CD in Indiana that did not trend hard to the Pubs in 2012, and no doubt over time it will continue to trend Dem, as the Pubs thin out in Marion County.

I think with good candidates the Dems could take and hold CDs 1, 7 (D+3.0), and 2 (R+0.7). They would be competitive and based on 2012 would carry 5 (R+3.3). They have outside chances in 9 (R+5) and 8 (R+6) which they might have taken in 2006 or 2008, but not in 2012.

I think it takes an intentional Dem gerrymander to do better. That requires separating Lake and Porter into separate CDs.

[Torie]

"I think with good candidates the Dems could take and hold CDs 1, 7 (D+3.0), and 2 (R+0.7). They would be competitive and based on 2012 would carry 5 (R+3.3). They have outside chances in 9 (R+5) and 8 (R+6) which they might have taken in 2006 or 2008, but not in 2012."

If you are using 2008 numbers for the PVI's, the Dems were at their absolute max outside of the NE corner of the state and Marion. So I don't think there is much hope for them with CD-05, much less CD's 8 and 9. One upon a time, both 8 and 9 were competitive. Those days are over now. So I think only CD-2 is in real play. If economic issues replace social and cultural and symbolic issues once again, then it will be a whole new world. Your political future would be considerably more rosy as it were. The current regime makes life in the burbs of most major metro areas outside the South a chore.

[Torie]

Here's Maryland. Remarkably enough, no township chops, except one by MD-06 into Montgomery County. I suspect Muon2 already drew this map to the tee. There is really but one way to do it, given his metrics. Anyway, MD-08 has a Pub PVI of about 1%, so it's a tossup, and the Pubs pick up 1.5 seats.

I don't think the VRA requires three black CD's. While it may be theoretically possible to draw two 50% BVAP CD's in the DC area, it's too marginal, and requires traveling chops and taking in large swaths of white areas. In addition, MD is 28% BVAP, so two CD's is their appropriate share of CD's in any event.  MD-06 is 41.8% BVAP, and voted 23.6% for McCain, and is 15.8% Hispanic, so I suspect the BVAP in Dem primaries is close to 50% anyway in MD-06, and thus is a performing black CD anyway, or at least it will be when Steny Hoyer retires. MD may be the only state where blacks are in a position to get more CD's than their allocable share.

[muon2]

Here's Maryland. Remarkably enough, no township chops, except one by MD-06 into Montgomery County. I suspect Muon2 already drew this map to the tee. There is really but one way to do it, given his metrics. Anyway, MD-08 has a Pub PVI of about 1%, so it's a tossup, and the Pubs pick up 1.5 seats.

I don't think the VRA requires three black CD's. While it may be theoretically possible to draw two 50% BVAP CD's in the DC area, it's too marginal, and requires traveling chops and taking in large swaths of white areas. In addition, MD is 28% BVAP, so two CD's is their appropriate share of CD's in any event.  MD-06 is 41.8% BVAP, and voted 23.6% for McCain, and is 15.8% Hispanic, so I suspect the BVAP in Dem primaries is close to 50% anyway in MD-06, and thus is a performing black CD anyway, or at least it will be when Steny Hoyer retires. MD may be the only state where blacks are in a position to get more CD's than their allocable share.

Interesting. We looked at MD quite a bit in 2013 with an eye to chops and erosity, but that was before we had metrics for UCCs and macrochops.

I wasn't aware of townships in MD, so I did a little research. The precinct numbers in DRA are divisions of historical election districts in each county. The county boards of election adjust precincts within them according to rules set by the state board of election. As in other states the municipalities are less numerous than the census places indicated in DRA. In any case the election districts don't correspond to the actual incorporated cities and sometimes they chop them. Election districts do cover the whole of each county. Should they be used in MD as the county subunits?

Back to UCCs, since we didn't look at them in our last analysis of MD. Both the Baltimore and Washington UCCs have a cover of 4 and pack of 3. I see that you have a cover penalty of 1 and pack penalty of 1 for Balto, and a pack penalty of 2 for Wash, along with 6 county chops for a chop score of 10.

This was one of my offerings from 2013. It has 6 county chops, one of which is a microchop since we ere giving credit for those back then. I count a cover penalty of 1 in Balto and a pack penalty of 1 in Wash for a chop score of 8. It has 2 BVAP majority CDs (5 and 7) and one at 41% (CD 4). The partisan breakdown is 5D, 1e (CD-3), 2R for a skew of 0.



I also discovered that we hadn't firmed up rules for ferries back then and this plan used the Smith Island ferry which is seasonal from the western shore. It has only 5 county chops, 1 pack penalty in Balto, and 1 cover and 1 pack penalty in Wash, so it also has a chop of 8. It has 3D, 2d (CD 2 is D+1.9), 2e (CDs 3 and 6), 1R for a skew of D+1. It would now be disallowed due to the seasonal ferry connection.

[Torie]

You are always always more skilled at this than I am aren't you? Anyway, to answer your question, the locality definitions to use need to match how the precincts are drawn. If the precincts as drawn do not nest into municipalities, than using that metric is impracticable. If the precincts nest in both historical election districts, and municipalities, and the election districts have no governmental function, other than the bookkeeping conventions of the county boards of election, than use municipalities. It's the same thing for villages in Michigan. If the precincts nested in them, then they should have meaning (even if only to limit the chopping of them to but one per CD chop), but if not, then they need to be ignored.

Here is the next state for you to embarrass me yet again - my favorite state next to Michgian - PA!.  

This one was an exercise is avoiding macrochops. Thus, the erose lines of the chop in Adams County to keep the chop number down just below 35,285.  Ditto for Blair County.  Muon2’s metrics necessitated the erose mess of the chop there, plus having to book an extra highway cut, as the highway south goes out of PA-12, and then back in again. If a macrochop of Blair was unavoidable, it might have paid, when it came to total highway cuts, to put Huntington County into PA-05, in order to reduce to a relatively small number the chop into Blair.  That would have made the map hideously ugly, but still reduced the total number of highway cuts, because each locality is assessed in the macrochopped county to see if it involves a state highway cut, and PA is blessed with a plethora of state highways.  (It’s not North Carolina, which seems to be more parsimonious in that department.)  No system is perfect, and every system involves game playing.  That was certainly the case here.

As to the partisan effect, it remains a two seat pickup for the Dems.  PA-15 moves from tossup to Dem with a Dem PVI of 3% (Dent would have won anyway in 2012, but this is an “on-paper” exercise), and PA-03 from Pub to tossup (Pub PVI of 1.16%).  The Dems might have picked up PA-03 in 2012, but the CD trended pretty hard to the Pubs in 2014 (maybe up to 2.5% Pub), and the Pubs probably would have taken it back in 2014. PA-07 moves from Pub to heavily Dem (Dem PVI of 6%). In 2014, however, PA-06 was one of the very few places in PA that trended Dem (Delaware County is getting more black), and PA-07 as drawn by the Pubs is now a tossup CD on paper, so that CD at the moment is but a half seat pickup on paper.  That is also true for PA-15, which in 2012 saw its Dem PVI crash down from 3%, to but 0.6% (meaning under current lines what has a tossup seat in 2008 is now a Pub seat). PA-06 was a tossup under the Pub gerrymander, and is still a tossup, so no change there. So PA-15 and PA-03 are half seat gains for the Dem, and PA-07 a full seat gain, thus totaling a two seat pickup for the Dems on paper.

This is yet another example of how even hideous gerrymanders, such as the Pub one in PA, really quite often didn’t net much for the Pubs: just PA-07 and maybe a two year “lease” of PA-03.  It’s all sound and fury - signifying not much.



Oh, never mind that map. I forgot about the cover and pack penalty with respect to Harrisburg. I realized that Cumberland County was way too populated and close to Harrisburg to not be part of a Harrisburg based UCC cluster, so I looked at Jimrtex's urban cluster map, and voila, it is. So that had to be dealt with, and in the course of things, the map actually improved, the saga of finessing macrochops came to an end, and a chop with respect to PA-10 was lost as well, so the map below has a chop score that improved by three points. Wow! Good luck beating this one Muon2.

[Adam Griffin]

My metric is actually much easier to compute than some of the ones proffered to judges in hopes of finding a justiciable standard.

What you want is my computation of SKEW.
1. Find the PVI for the state as a fraction (or divide the percent by 100) and multiply that by 4 times the number of districts. This is the expected excess in the delegation for the party in a neutral map. Count Republicans as a negative number.
2. Find the PVI's of the actual districts. Count 0 for each highly competitive district (PVI 0 or 1), +1 for all other Democratic districts, and -1 for all other Republican districts. The total is the expected excess in the delegation under the actual map.
3. Take the number from step 2 and subtract the number from step 1. Express a negative number as a positive number in favor of the Republicans. The resulting positive number is the SKEW score, and lower numbers are closer to the ideal partisan fairness.
2A. You can substitute the actual delegation for the hypothetical delegation based on PVIs.

AZ is R+7. 4 times that percentage is -28%, and times 9 seats is -2.52 (rounded to -3), which is the neutral expectation. The map has 1 highly competitive CD (AZ-9), 2 Dem CDs (AZ-3,7) and 6 Pub CDs for a partisan expectation of -4. The SKEW is -1 or R+1. Over the decade it should slightly favor the Pubs.

The actual delegation is 4D, 5R or an actual excess of -1. Compared to -3 that gives a current SKEW of D+2, which gets howls from the Pubs. But unless the district PVI's are way off then either the Dems are overperforming or the Pubs are underperforming the map.

edit: Based on 2012 results only the delegation is 5D 4R so the skew is D+4.

Perhaps it has already been answered, but how do you define "competitive"?

[muon2]

My metric is actually much easier to compute than some of the ones proffered to judges in hopes of finding a justiciable standard.

What you want is my computation of SKEW.
1. Find the PVI for the state as a fraction (or divide the percent by 100) and multiply that by 4 times the number of districts. This is the expected excess in the delegation for the party in a neutral map. Count Republicans as a negative number.
2. Find the PVI's of the actual districts. Count 0 for each highly competitive district (PVI 0 or 1), +1 for all other Democratic districts, and -1 for all other Republican districts. The total is the expected excess in the delegation under the actual map.
3. Take the number from step 2 and subtract the number from step 1. Express a negative number as a positive number in favor of the Republicans. The resulting positive number is the SKEW score, and lower numbers are closer to the ideal partisan fairness.
2A. You can substitute the actual delegation for the hypothetical delegation based on PVIs.

AZ is R+7. 4 times that percentage is -28%, and times 9 seats is -2.52 (rounded to -3), which is the neutral expectation. The map has 1 highly competitive CD (AZ-9), 2 Dem CDs (AZ-3,7) and 6 Pub CDs for a partisan expectation of -4. The SKEW is -1 or R+1. Over the decade it should slightly favor the Pubs.

The actual delegation is 4D, 5R or an actual excess of -1. Compared to -3 that gives a current SKEW of D+2, which gets howls from the Pubs. But unless the district PVI's are way off then either the Dems are overperforming or the Pubs are underperforming the map.

edit: Based on 2012 results only the delegation is 5D 4R so the skew is D+4.

Perhaps it has already been answered, but how do you define "competitive"?

I use three categories of PVI. The groupings were based on congressional electoral results over a decade. Highly competitive districts have a PVI of 0 or 1 and statistically have an equal chance of going to either party in an election. Competitive districts have a PVI of 2 to 5 and favor the likely party by about a 3 to 1 in election outcomes. Uncompetitive districts have a PVI of 6 or more and go to the favored party over 90% of the time.

For SKEW the competitive and uncompetitive districts are lumped together since those are districts that clearly favor a party. They all count 1 towards the favored party.

For POLARIZATION the competitive districts count 1 each and the uncompetitive districts count 2 each. Lower scores indicate a plan where the seats will be more responsive to swings in the electorate.

[Torie]

My metric is actually much easier to compute than some of the ones proffered to judges in hopes of finding a justiciable standard.

What you want is my computation of SKEW.
1. Find the PVI for the state as a fraction (or divide the percent by 100) and multiply that by 4 times the number of districts. This is the expected excess in the delegation for the party in a neutral map. Count Republicans as a negative number.
2. Find the PVI's of the actual districts. Count 0 for each highly competitive district (PVI 0 or 1), +1 for all other Democratic districts, and -1 for all other Republican districts. The total is the expected excess in the delegation under the actual map.
3. Take the number from step 2 and subtract the number from step 1. Express a negative number as a positive number in favor of the Republicans. The resulting positive number is the SKEW score, and lower numbers are closer to the ideal partisan fairness.
2A. You can substitute the actual delegation for the hypothetical delegation based on PVIs.

AZ is R+7. 4 times that percentage is -28%, and times 9 seats is -2.52 (rounded to -3), which is the neutral expectation. The map has 1 highly competitive CD (AZ-9), 2 Dem CDs (AZ-3,7) and 6 Pub CDs for a partisan expectation of -4. The SKEW is -1 or R+1. Over the decade it should slightly favor the Pubs.

The actual delegation is 4D, 5R or an actual excess of -1. Compared to -3 that gives a current SKEW of D+2, which gets howls from the Pubs. But unless the district PVI's are way off then either the Dems are overperforming or the Pubs are underperforming the map.

edit: Based on 2012 results only the delegation is 5D 4R so the skew is D+4.

Perhaps it has already been answered, but how do you define "competitive"?

A PVI that does not exceed 1.5% in favor of either party. I am not sure if exactly 1.5% is deemed competitive or not. I deem it competitive, but I guess it does not make any difference. Since only one vote will be involved either way, the odds of it being exactly 1.5% are de minimus.

That is for purposes of calculating the partisan skew, which is how much a variance the parties' share of CD's varies from the overall popular vote. What does not affect the skew, but is noted in passing, are lean seats, which have a PVI of no more than 5.5%. Sometimes the minority party wins such seats, but not often. Muon2 when he puts up the partisan split of maps uses "R" or "D" for seats with a PVI in excess of 5.5%, and "r" or "d" for seats with a PVI of more than 1.5%, but not in excess of 5.5%, and "e" for the competitive or tossup seats.

Based on the above defined categories, my last map above has 8R, 0r, 3e, 2d and 5D. So the Pubs have 1 more seat than the Dems, 8-7. PA had a PVI of 1.53% Dem in 2008. You double that figure to 3.06% and add 50%, which sums to 53.06%. 53.06% times 18 seats equals 9.55 seats, which is the "unskewed" share of Dem seats in PA, with the Pubs having 8.45 seats. So the Dems with no skew would have 1.1 seats more than the Pubs (rounded to 1 as the nearest whole integer), with 8 Dem seats, and 7 Pub seats. Instead they have 1 seat less, so the SKEW is 1. That is the way I calculate it anyway.

I didn't know Muon2 had a POLARIZATION score. That gives some meaning to r and d seats. My POLARIZATION score is ((8+5)*(2)) + 2 = 28. Well, PA is a polarized state. But it least it has three "tossup" seats, or did in 2008. It still does I guess, with PA-03 no longer a tossup, but PA-15 becoming a tossup seat, which means the SKEW is increasing however. Oh, PA-06 is no longer a tossup seat either, but a Pub seat (the Pub trend in Chester and Berks counties was very strong in 2012). So now the score is 10-6. The PA PVI for 2012 was 0.78%, which when doubled plus 50% is 51.56% or 9.28 seats Dem seats and 8.72 Pub seats, which is 0.56 more Dem seats than Pub seats, or when rounded to the nearest whole integer, 1 more seat than the Pubs, or 8 Dem seats and 7 Pub seats. So now the Dems are 3 seats short, which means the SKEW is now 3 in favor of the Pubs. If PA had trended just a tad more to the Pubs, the SKEW would have been 2 (with each party with no SKEW having an equal number of seats).

[muon2]

You are always always more skilled at this than I am aren't you? Anyway, to answer your question, the locality definitions to use need to match how the precincts are drawn. If the precincts as drawn do not nest into municipalities, than using that metric is impracticable. If the precincts nest in both historical election districts, and municipalities, and the election districts have no governmental function, other than the bookkeeping conventions of the county boards of election, than use municipalities. It's the same thing for villages in Michigan. If the precincts nested in them, then they should have meaning (even if only to limit the chopping of them to but one per CD chop), but if not, then they need to be ignored.

County subunits will tend to be unique to each state. MI has townships and incorporated cities that clearly delineate where local governments lie. The villages in MI are really adjuncts of the township in terms of their governance. PA has local government associated with their townships and boroughs, too.

MD has incorporated cities, towns and villages with local government. There are also places, colloquially referred to as towns, that are not towns and lack local government. Some of these like Columbia and Germantown are more populous than any incorporated city except Baltimore. The election districts seem to be relatively static (maybe a MD expert can weigh in) so they could be reasonable subunits, but they don't correspond to units of government that residents would recognize. It would mean that divisions with no CoI basis become the standard in MD, unlike in MI or PA.

[Torie]

You are always always more skilled at this than I am aren't you? Anyway, to answer your question, the locality definitions to use need to match how the precincts are drawn. If the precincts as drawn do not nest into municipalities, than using that metric is impracticable. If the precincts nest in both historical election districts, and municipalities, and the election districts have no governmental function, other than the bookkeeping conventions of the county boards of election, than use municipalities. It's the same thing for villages in Michigan. If the precincts nested in them, then they should have meaning (even if only to limit the chopping of them to but one per CD chop), but if not, then they need to be ignored.

County subunits will tend to be unique to each state. MI has townships and incorporated cities that clearly delineate where local governments lie. The villages in MI are really adjuncts of the township in terms of their governance. PA has local government associated with their townships and boroughs, too.

MD has incorporated cities, towns and villages with local government. There are also places, colloquially referred to as towns, that are not towns and lack local government. Some of these like Columbia and Germantown are more populous than any incorporated city except Baltimore. The election districts seem to be relatively static (maybe a MD expert can weigh in) so they could be reasonable subunits, but they don't correspond to units of government that residents would recognize. It would mean that divisions with no CoI basis become the standard in MD, unlike in MI or PA.

Yes, but it is practicable. It again is one of those situations where a nickel's worth of perfection costs a pound of headache. The main thing really for these little rules, is to have random effects. As long as it is random, there is no gerrymander. That is the beauty of the system. That is why it is OK to have bridge chops per my metrics. The effects will be random as to what party it favors. The only issue is how much of a partisan variance different maps have that are high scoring. If MD wants the metrics to match communities of interest when it comes to localities, the fix is to fix the precinct boundaries to nest into localities that have some governmental jurisdictional meaning of more import, aka in this context called "communities of interest."  Absent that, I think it appropriate to live with spending the nickel to avoid losing the pound.

[muon2]

You are always always more skilled at this than I am aren't you? Anyway, to answer your question, the locality definitions to use need to match how the precincts are drawn. If the precincts as drawn do not nest into municipalities, than using that metric is impracticable. If the precincts nest in both historical election districts, and municipalities, and the election districts have no governmental function, other than the bookkeeping conventions of the county boards of election, than use municipalities. It's the same thing for villages in Michigan. If the precincts nested in them, then they should have meaning (even if only to limit the chopping of them to but one per CD chop), but if not, then they need to be ignored.

County subunits will tend to be unique to each state. MI has townships and incorporated cities that clearly delineate where local governments lie. The villages in MI are really adjuncts of the township in terms of their governance. PA has local government associated with their townships and boroughs, too.

MD has incorporated cities, towns and villages with local government. There are also places, colloquially referred to as towns, that are not towns and lack local government. Some of these like Columbia and Germantown are more populous than any incorporated city except Baltimore. The election districts seem to be relatively static (maybe a MD expert can weigh in) so they could be reasonable subunits, but they don't correspond to units of government that residents would recognize. It would mean that divisions with no CoI basis become the standard in MD, unlike in MI or PA.

Yes, but it is practicable. It again is one of those situations where a nickel's worth of perfection costs a pound of headache. The main thing really for these little rules, is to have random effects. As long as it is random, there is no gerrymander. That is the beauty of the system. That is why it is OK to have bridge chops per my metrics. The effects will be random as to what party it favors. The only issue is how much of a partisan variance different maps have that are high scoring. If MD wants the metrics to match communities of interest when it comes to localities, the fix is to fix the precinct boundaries to nest into localities that have some governmental jurisdictional meaning of more import, aka in this context called "communities of interest."  Absent that, I think it appropriate to live with spending the nickel to avoid losing the pound.

Why should precinct boundaries matter other than that its the smallest unit on DRA? DRA won't exist for 2020 as it stands now (the software it's based on is no longer supported by Microsoft). Many states require that precincts be adjusted after each decade's remap to better reflect the new district boundaries. That seems like good policy since it cuts down on costs associated with ballot types and allows more balanced numbers of voters at a polling place.

[Torie]

You are always always more skilled at this than I am aren't you? Anyway, to answer your question, the locality definitions to use need to match how the precincts are drawn. If the precincts as drawn do not nest into municipalities, than using that metric is impracticable. If the precincts nest in both historical election districts, and municipalities, and the election districts have no governmental function, other than the bookkeeping conventions of the county boards of election, than use municipalities. It's the same thing for villages in Michigan. If the precincts nested in them, then they should have meaning (even if only to limit the chopping of them to but one per CD chop), but if not, then they need to be ignored.

County subunits will tend to be unique to each state. MI has townships and incorporated cities that clearly delineate where local governments lie. The villages in MI are really adjuncts of the township in terms of their governance. PA has local government associated with their townships and boroughs, too.

MD has incorporated cities, towns and villages with local government. There are also places, colloquially referred to as towns, that are not towns and lack local government. Some of these like Columbia and Germantown are more populous than any incorporated city except Baltimore. The election districts seem to be relatively static (maybe a MD expert can weigh in) so they could be reasonable subunits, but they don't correspond to units of government that residents would recognize. It would mean that divisions with no CoI basis become the standard in MD, unlike in MI or PA.

Yes, but it is practicable. It again is one of those situations where a nickel's worth of perfection costs a pound of headache. The main thing really for these little rules, is to have random effects. As long as it is random, there is no gerrymander. That is the beauty of the system. That is why it is OK to have bridge chops per my metrics. The effects will be random as to what party it favors. The only issue is how much of a partisan variance different maps have that are high scoring. If MD wants the metrics to match communities of interest when it comes to localities, the fix is to fix the precinct boundaries to nest into localities that have some governmental jurisdictional meaning of more import, aka in this context called "communities of interest."  Absent that, I think it appropriate to live with spending the nickel to avoid losing the pound.

Why should precinct boundaries matter other than that its the smallest unit on DRA? DRA won't exist for 2020 as it stands now (the software it's based on is no longer supported by Microsoft). Many states require that precincts be adjusted after each decade's remap to better reflect the new district boundaries. That seems like good policy since it cuts down on costs associated with ballot types and allows more balanced numbers of voters at a polling place.

As long as the precinct boundaries are fixed, fine with me, or the state is willing to fix them after the fact. Since we are talking about enacting state laws here, if the metric is to follow municipalities, part and parcel of the legislation, would be to fix the precinct boundaries, so I guess it is a moot point. For purposes of the game I am playing however, it is a pain in the butt to take cognizance of something which does not match the precincts, or when it comes to your city neighborhoods concept, no boundaries at all.

[jimrtex]

You are always always more skilled at this than I am aren't you? Anyway, to answer your question, the locality definitions to use need to match how the precincts are drawn. If the precincts as drawn do not nest into municipalities, than using that metric is impracticable. If the precincts nest in both historical election districts, and municipalities, and the election districts have no governmental function, other than the bookkeeping conventions of the county boards of election, than use municipalities. It's the same thing for villages in Michigan. If the precincts nested in them, then they should have meaning (even if only to limit the chopping of them to but one per CD chop), but if not, then they need to be ignored.

County subunits will tend to be unique to each state. MI has townships and incorporated cities that clearly delineate where local governments lie. The villages in MI are really adjuncts of the township in terms of their governance. PA has local government associated with their townships and boroughs, too.

MD has incorporated cities, towns and villages with local government. There are also places, colloquially referred to as towns, that are not towns and lack local government. Some of these like Columbia and Germantown are more populous than any incorporated city except Baltimore. The election districts seem to be relatively static (maybe a MD expert can weigh in) so they could be reasonable subunits, but they don't correspond to units of government that residents would recognize. It would mean that divisions with no CoI basis become the standard in MD, unlike in MI or PA.

Yes, but it is practicable. It again is one of those situations where a nickel's worth of perfection costs a pound of headache. The main thing really for these little rules, is to have random effects. As long as it is random, there is no gerrymander. That is the beauty of the system. That is why it is OK to have bridge chops per my metrics. The effects will be random as to what party it favors. The only issue is how much of a partisan variance different maps have that are high scoring. If MD wants the metrics to match communities of interest when it comes to localities, the fix is to fix the precinct boundaries to nest into localities that have some governmental jurisdictional meaning of more import, aka in this context called "communities of interest."  Absent that, I think it appropriate to live with spending the nickel to avoid losing the pound.

Because of the lack of incorporated places in Maryland, census designated places, may be more formally recognized as COI than in other states. One should not assume that CDP are imposed on a top down basis by the federal government. Plumas County California has a large number of CDP's (46 for about 20,000 people) which were recognized to help in the redistricting of supervisor districts. Miami-Dade Florida formally recognizes census designated places as the basis for its community councils, which are elected bodies. 6 of the original 16 have been disbanded as areas have incorporated.

Maryland Department of Planning Census Designated Places

It would be expected that subareas of counties would be delineated prior to the census.

I understand that for the purpose of your immediate study that you might choose some arbitrary COI such as election precincts, but this is contrary to Muon2's methodology.

[muon2]

I appreciate the effort to see how much swing there is based on the application of the rules. I think it's notable that our various MD maps produced about the same swing, that's a good sign IMO.

As part of a system, I struggle with the delineation of county subunits. 25 of the states have clear subunits in the form of minor civil divisions used by the Census, they are easy. The other 25 do not have MCDs that are ready to use. The Census has county subdivisions in each state, including those without MCDs, so I'm prone to use those. In MD that matches up with the election districts you used, so I think we're all happy. That isn't always the case such as in FL, where the ongoing litigation has requirements about incorporated municipalities.

In the end I'd like a table that for each state it specifies what subunits are used. Precincts can be useful for this type of test, but they aren't reliable as fixed entities.

[Adam Griffin]

Thanks for the answers - and wow, I'm an idiot - you had already answered that (at least in the primary way I needed to know)...in the post I quoted, no less. I appreciate the additional info.

[muon2]

My metric is actually much easier to compute than some of the ones proffered to judges in hopes of finding a justiciable standard.

What you want is my computation of SKEW.
1. Find the PVI for the state as a fraction (or divide the percent by 100) and multiply that by 4 times the number of districts. This is the expected excess in the delegation for the party in a neutral map. Count Republicans as a negative number.
2. Find the PVI's of the actual districts. Count 0 for each highly competitive district (PVI 0 or 1), +1 for all other Democratic districts, and -1 for all other Republican districts. The total is the expected excess in the delegation under the actual map.
3. Take the number from step 2 and subtract the number from step 1. Express a negative number as a positive number in favor of the Republicans. The resulting positive number is the SKEW score, and lower numbers are closer to the ideal partisan fairness.
2A. You can substitute the actual delegation for the hypothetical delegation based on PVIs.

AZ is R+7. 4 times that percentage is -28%, and times 9 seats is -2.52 (rounded to -3), which is the neutral expectation. The map has 1 highly competitive CD (AZ-9), 2 Dem CDs (AZ-3,7) and 6 Pub CDs for a partisan expectation of -4. The SKEW is -1 or R+1. Over the decade it should slightly favor the Pubs.

The actual delegation is 4D, 5R or an actual excess of -1. Compared to -3 that gives a current SKEW of D+2, which gets howls from the Pubs. But unless the district PVI's are way off then either the Dems are overperforming or the Pubs are underperforming the map.

edit: Based on 2012 results only the delegation is 5D 4R so the skew is D+4.

Perhaps it has already been answered, but how do you define "competitive"?

A PVI that does not exceed 1.5% in favor of either party. I am not sure if exactly 1.5% is deemed competitive or not. I deem it competitive, but I guess it does not make any difference. Since only one vote will be involved either way, the odds of it being exactly 1.5% are de minimus.

That is for purposes of calculating the partisan skew, which is how much a variance the parties' share of CD's varies from the overall popular vote. What does not affect the skew, but is noted in passing, are lean seats, which have a PVI of no more than 5.5%. Sometimes the minority party wins such seats, but not often. Muon2 when he puts up the partisan split of maps uses "R" or "D" for seats with a PVI in excess of 5.5%, and "r" or "d" for seats with a PVI of more than 1.5%, but not in excess of 5.5%, and "e" for the competitive or tossup seats.

Based on the above defined categories, my last map above has 8R, 0r, 3e, 2d and 5D. So the Pubs have 1 more seat than the Dems, 8-7. PA had a PVI of 1.53% Dem in 2008. You double that figure to 3.06% and add 50%, which sums to 53.06%. 53.06% times 18 seats equals 9.55 seats, which is the "unskewed" share of Dem seats in PA, with the Pubs having 8.45 seats. So the Dems with no skew would have 1.1 seats more than the Pubs (rounded to 1 as the nearest whole integer), with 8 Dem seats, and 7 Pub seats. Instead they have 1 seat less, so the SKEW is 1. That is the way I calculate it anyway.

I didn't know Muon2 had a POLARIZATION score. That gives some meaning to r and d seats. My POLARIZATION score is ((8+5)*(2)) + 2 = 28. Well, PA is a polarized state. But it least it has three "tossup" seats, or did in 2008. It still does I guess, with PA-03 no longer a tossup, but PA-15 becoming a tossup seat, which means the SKEW is increasing however. Oh, PA-06 is no longer a tossup seat either, but a Pub seat (the Pub trend in Chester and Berks counties was very strong in 2012). So now the score is 10-6. The PA PVI for 2012 was 0.78%, which when doubled plus 50% is 51.56% or 9.28 seats Dem seats and 8.72 Pub seats, which is 0.56 more Dem seats than Pub seats, or when rounded to the nearest whole integer, 1 more seat than the Pubs, or 8 Dem seats and 7 Pub seats. So now the Dems are 3 seats short, which means the SKEW is now 3 in favor of the Pubs. If PA had trended just a tad more to the Pubs, the SKEW would have been 2 (with each party with no SKEW having an equal number of seats).

The official PVI is the average from the last two presidential years. So for PA that is D+1.16%. I multiply that by 4 and the number of CDs to get 0.83 or D+1 seat. That's the same math that Torie used without the extra steps to add then subtract the 50%:

(PVI*2) + 0.5 - {1 - [(PVI*2)+0.5]} = (PVI*2) + 0.5 - 1 + (PVI*2) +0.5 = (PVI*4)

Torie is correct that a skewless delegation in PA would be 8D, 7R, 1 even.

We missed you during the Forum Redistricting Commission a year ago where I walked through all my scoring rules one at a time. Unfortunately it stalled during the details about how to treat the independent cities in VA.

[Nyvin]

This was one of my offerings from 2013. It has 6 county chops, one of which is a microchop since we ere giving credit for those back then. I count a cover penalty of 1 in Balto and a pack penalty of 1 in Wash for a chop score of 8. It has 2 BVAP majority CDs (5 and 7) and one at 41% (CD 4). The partisan breakdown is 5D, 1e (CD-3), 2R for a skew of 0.

This would be a good map for Maryland honestly,  just give the GOP their darn northwest seat and be done with it.   One less thing for them to deflect on with the red state gerrymanders. 

If places like Texas and North Carolina had fair maps I'd be in favor of this. 

[Adam Griffin]

Well, I thought I'd try my first rough-shot at this without fully understanding all of the components. I'm sure I can do better.

With the requirement of 4 VRA districts, this significantly impacts drawing any Georgia map that adheres to the two characteristics desired above. Furthermore, considering Georgia doesn't actually value municipal boundaries in terms of precinct design more often than not...well, what can you do?

But hey, we don't really value any of that around here. At least I didn't use point contiguity!

Assuming that Georgia is R+6, this map is 8R & 6D, with one of those D districts appearing to be competitive at D+1 (CD11). That's based solely on my familiarity with precinct swings in the northern swing precincts of Fulton & Dekalb, as well as those in portions of Cobb & Gwinnett (it was a D+3 district in 2008, but vast segments of this area at the precinct level saw Romney do 5-6 points better than McCain - among some of the largest swings in the state).

CD3 (which would pit Buddy Carter & Rick Allen against one another) is D+3 but everything else is obviously quite safe. The next closest districts would be Sanford Bishop's CD2 at D+8 and the now-congressmanless CD4 at R+9.

CD2 in particular had to be especially irregular because such is necessary just to get it to 50.2% black VAP. In contrast, the Bishop's real-life 2011 district was only 49.4% black VAP. The other 3 VRA districts are between 54.5% and 56.3% black VAP.

[Sol]

There are a few helpful tricks in drawing Georgia. For example, DeKalb is the perfect size for one district.

[Adam Griffin]

There are a few helpful tricks in drawing Georgia. For example, DeKalb is the perfect size for one district.

Right. Cobb's another good one, but Dekalb can be used to generate a wholly-contained VRA district. However (and I am not 100% sure about this), putting Dekalb entirely to itself given the nature of white liberals in both Fulton and Dekalb would likely make it very difficult to draw 5 solid D seats + a competitive 6th that wasn't fragmented and snake-like. Additionally, it might wreak havoc on the shape of the other VRA districts.

[Adam Griffin]

There are a few helpful tricks in drawing Georgia. For example, DeKalb is the perfect size for one district.

Right. Cobb's another good one, but Dekalb can be used to generate a wholly-contained VRA district. However (and I am not 100% sure about this), putting Dekalb entirely to itself given the nature of white liberals in both Fulton and Dekalb would likely make it very difficult to draw 5 solid D seats + a competitive 6th that wasn't fragmented and snake-like. Additionally, it might wreak havoc on the shape of the other VRA districts.

So I was wrong about the 5. This is another possibility using Dekalb as its own CD, which does make it possible to generate a comparatively-competitive CD (same numbers for 2008; 54-44 Obama) while maintaining 3 VRA districts in the metro, but that competitive district is a mess in terms of geography.

EDIT: You can tidy it up a bit more on the southern end, but this makes the purple district only 50.3% black VAP.

[Sol]

This would be the alternative I would put regarding Torie's map (outside of Metro Atlanta, the map is the same). Unlike his, it retains three black VAP majority districts in the Atlanta area. I don't recall a erosity connection map for Georgia; obviously the black CD is quite erose--I drew it that way to avoid a UCC cover penalty--at least as I understand it. Depending on how the penalties are assessed for both, it might be preferable to draw a seat based in Coweta/Spalding counties and adjacent non-Atlanta areas and then a similar arrangement with Newton/Rockdale/Walton/Gwinnett. All of the Atlanta districts are within 1000 of the ideal.

[Torie]

If one can draw a compact 50% BVAP CD, that triggers the VRA. But the CD need not by 50% BVAP. It just needs to be a performing district, and the test for that is whether of not the black voters make up enough of a percentage of voters in the Dem primary to nominate a black, which might be as low as 40% BVAP or even lower, depending on how Hispanic a CD is, and how susceptible whites in a Dem primary are to voting for a black. Only 3 black CD's need be drawn in Georgia. The SW corner does not have enough blacks to draw a compact 50% BVAP CD. So it is a two step process, first the 50% test, and if met, then the performing test.