Back to questioning the Muon2 rules

[Torie]

I drew this Pub gerrymander circa after the 2020 census for the benefit of Krazen, to display to him the joys of at once legal yet aesthetic Pub gerrymanders (yes, I know I was soft in Pubbing up MI-03 (shifting a bit of Pub strength into MI-03 from MI-02), but my heart just wasn't in it to squeeze out another Pub point or two, at the cost of just going slash and burn), and it raised an issue that we have discussed before. Below is my macro-chop into Macomb. I take it I could avoid two erosity penalty points by avoiding having the chop into Macomb result in a CD line touching either Centerline or Roseville. Is that wise? Remind me what the policy reason here is again. By avoiding the touchings, the lines would become more jagged (it would look like a staircase), leaving the "artist" community very upset at the discretion of it all. We strive for straight lines, unless the cost is too high in the gerrymander mode, not jagged ones.



Feel free to move this post elsewhere, and delete this thread if you like, Muon2. I am just too lazy to search for the old thread that included our discussions. No doubt you know exactly where it is!

[muon2]

You have macrochopped Warren and the erosity is much higher than you think. Once it is macrochopped we consider the subunits of the macrochop. In this case it would be the precincts within Warren. The erosity just across the two parts of Warren is something like 13 or 14 given the number of precincts that are connected by road there. If you try to avoid touching Center Line it will only make the erosity that much worse through the very stair steps you describe.

As I look at it, the line across Warren isn't that straight. The better plan would fill that last precinct surrounding Center Line with blue Warren, then shift the northwest blue precinct over to pink Warren. Still both are gross compared to any plan that keeps Warren whole or at least to a simple chop.

[Torie]

You have macrochopped Warren and the erosity is much higher than you think. Once it is macrochopped we consider the subunits of the macrochop. In this case it would be the precincts within Warren. The erosity just across the two parts of Warren is something like 13 or 14 given the number of precincts that are connected by road there. If you try to avoid touching Center Line it will only make the erosity that much worse through the very stair steps you describe.

As I look at it, the line across Warren isn't that straight. The better plan would fill that last precinct surrounding Center Line with blue Warren, then shift the northwest blue precinct over to pink Warren. Still both are gross compared to any plan that keeps Warren whole or at least to a simple chop.

You do it precinct by precinct? That seems insane to me (ask me if you want, as to why). I thought it was by subunit by subunit, which makes more sense, putting aside the issue that I raised. It is not easy to do a simple chop of Warren. But yeah, it could be better, if not in the gerrymandering mode, when it comes to straight lines probably.

I guess the real issue is the zoom within zoom. First you zoom into a macro chopped county, and then you zoom within a macro chopped subunit (this time precinct by precinct). I never understood that before. Maybe that has merit. Is that in your rules? If so, is the double zoom clearly stated? 

The result is that one would never unless one has to, macro-chop a subunit. Maybe given that Warren does not have hoods (ala Detroit), you are just reverting to precincts. That seems problematical. When does a subunit get large enough, that you search for hoods, and if you find none, revert to precincts?

[krazen1211]

That is not a bad map, good sir.

I would consider a 10th district that includes Macomb + Lapeer Counties, and places all of Warren into what you have as the 9th district. I might also give consideration to a map that puts Ingham County entirely in the 5th district and splits Genessee County instead.

[muon2]

A macrochop is 5% of the population of the district so any unit under 10% of the district can't be macrochopped. Warren in 2010 was 134K or 19% of a whole district of 706K so it can be macrochopped. When we divided Detroit they were in subunits that varied from 53K to 88K, and though the largest could be macrochopped, most chops would not exceed the 35K threshold for a macrochop. If the a Detroit subunit was macrochopped then we went to precincts. So Warren is closer to being like two Detroit neighborhoods.

In other jurisdictions we have looked for governmentally-designated areas that are typically between 5-10% of a CD. If there is a agreed upon subunit for cities over 10% of a CD then we would use that. If there isn't then the default is the precinct.

[Torie]

Is that all clearly stated in your rules? Using precincts I think is a terrible idea. They are too small among other things. And they are subject to arbitrary adjustment to affect the score. For example, knowing the Muon2 rules, as a Pub gerrymanderer, I would call up the Pub hack who runs the Board of Elections in Warren, and tell him to redraw the precinct lines as follows (each square is a precinct, with the CD border running between two precincts that stretch the length of the city of Warren):



Warren does I see have school districts in which the precincts conveniently enough seem to be nested, so I guess those could be used.

I see that the precinct lines have changed since 2008, but an approximation of school district nesting would look like the below. It isn’t very pretty, because the school district lines are not pretty.  



Of course, one can lose the subunit macro-chop by doing the below.  But then, if one is in the gerrymandering mode, one isn’t playing by the Muon2 rules anyway. Perhaps there should be chapter on the Muon2 rules for gerrymanders, which penalizes macro-chops. Putting aside the Warren contretemps, notice that my map avoids macro-chops, except where absolutely necessary to get the job done (i.e., Ingram).  That is not an accident, but I digress.  

[muon2]

The preexisting political lines are not expected to be pretty, but they do recognize real jurisdictions. The important idea is to have a previously agreed hierarchy of subunits for each state - from county down to census block.

The hierarchy in MI looked like:
County
City/Township
Planning Neighborhoods (Detroit only)
Precinct
Census Block (not usable in DRA)

The hierarchy in WA looked like
County
City/School District (approximate in DRA)
City-defined Neighborhoods (Seattle only)
Precinct
Census Block (not usable in DRA)

School districts were used in WA because there was no township government that completely covered and divided up all the unincorporated areas in a county. I'm not sure that school districts make the most sense for units like Warren. If there really is a need for a layer between city and precinct for cities in excess of 10% of a CD, I think zip codes would be better here.

In any case, it seems that what you were trying to do in the OP was a political gerrymander. The rules are supposed to make that task hard by penalizing plans that may look reasonable, but in fact chop up locales to achieve political goals. If you want to avoid the macrochop (35K and up) why not put Warren and Center Line in the blue CD minus a nice rectangular chop out of the NE corner of Warren for the pink?

[Torie]

A macro chop is 38k+. 

"If you want to avoid the macrochop (35K and up) why not put Warren and Center Line in the blue CD minus a nice rectangular chop out of the NE corner of Warren for the pink?"

Because Eastpointe is where the Dems are in highest concentration (this map is designed to be a "good government" gerrymander, following the pigs get fat, but hogs get slaughtered metric), and also the blacks for VRA purposes (with southern Warren next best). The two black CD's are on the cusp. The one going into Oakland is 49.9% BVAP, and the cyan one going into Macomb is 51% BVAP. That is based on the 2010 census. I suspect the percentages are lower now (and the map assumed the population changes were even across the counties, and they certainly are not in Wayne, Macomb or Oakland, so the cyan CD will need to dig deeper into Macomb anyway, and ditto for the black CD having to go deeper into Oakland. That will reduce the black percentages too.

[muon2]

I forgot the 2020 adjustment to the macrochop threshold.

The courts have been moving in the direction of allowing districts that can elect the candidate of choice even when there is not 50% BVAP. The 7th circuit was ok with that in IL legislative districts in the 2010 cycle (46% in one case). In general SCOTUS doesn't like bright lines in redistricting cases since they fear parties will figure out how to gerrymander around any safe harbor they provide. The VA case was a clear example of that softening line. I think that a 48% BVAP in MI will probably be ok in 2020 unless the Trump picks for SCOTUS totally redo the interpretation of performing minority districts in the next remap cycle. It seems more likely to me that they will move to greater deference to the legislature, not tighter requirements on minority districts.

Even if there is a move to tighter requirements for minority CDs, I think the rules are ok. Every plan will have the same constraint to meet the VRA and if that means chops are needed, the plan with the most judicious chops under the rules should be rewarded. It just may be that a partisan Pub plan won't get that reward.

[Torie]

Yes, I know, a lower percentage might well be legal. But politics is so racially polarized, particularly in Macomb, that the BVAP cannot go too much below 50% without it being problematical (bearing in mind that there are few Hispanics to push the CBVAP up). I suspect that in 2021 or 2022, when the lines are drawn, we might be down to about 45% BVAP for both CD's, if the BVAP is evenly split between them.

[jimrtex]

The preexisting political lines are not expected to be pretty, but they do recognize real jurisdictions. The important idea is to have a previously agreed hierarchy of subunits for each state - from county down to census block.

The hierarchy in MI looked like:
County
City/Township
Planning Neighborhoods (Detroit only)
Precinct
Census Block (not usable in DRA)

The hierarchy in WA looked like
County
City/School District (approximate in DRA)
City-defined Neighborhoods (Seattle only)
Precinct
Census Block (not usable in DRA)

School districts were used in WA because there was no township government that completely covered and divided up all the unincorporated areas in a county. I'm not sure that school districts make the most sense for units like Warren. If there really is a need for a layer between city and precinct for cities in excess of 10% of a CD, I think zip codes would be better here.

In any case, it seems that what you were trying to do in the OP was a political gerrymander. The rules are supposed to make that task hard by penalizing plans that may look reasonable, but in fact chop up locales to achieve political goals. If you want to avoid the macrochop (35K and up) why not put Warren and Center Line in the blue CD minus a nice rectangular chop out of the NE corner of Warren for the pink?

I think he may be getting into trouble with the VRA here. Between 2000 and 2010, the black population of Warren increased from 2.7% to 13.5%, presumably concentrated in the southern part of the city.

Three of the five city council districts are north-south strips. The remaining two are more of a north and south alignment, likely in the part of Center Line blocking a pure strip.

In 2000, Warren had the longest average tenure of any city over 100,000 in the US, at 35.5 years. It also had the oldest population, outside Florida and Hawaii.

Warren doubled in population every decade between 1940 and 1970 going from 20K to 180K, and has been sliding since. It appears that the population that moved in after WWII is aging out (someone who was 25 in 1970, is 72 now. The children have moved further north, or northwest, or way south (like Texas).