Local vs regional road connections

[muon2]

Torie has been heavily advocating for SKEW reducing plans of late and it's an important concern. I've reviewed our MI scoring from a year ago and the thread is worth a reread (it's only three pages long). The exercise was heavily data driven with tables and some regression analysis applied to the scoring metrics.

One of the things we did was address train's strong concern that inequality matter. I share some of that concern and the solution was to add INEQUALITY to CHOP for that arm of the Pareto test. It provided that a chop was ok if it was making the inequality substantially better. We already were applying the MI rule against multiple shared chops between pairs of districts.

If the goal is to not penalize one-bite chops that lower SKEW, then the obvious solution seems to just add SKEW to CHOP. Couple that with the MI rule that no two districts can share chops with more than one subunit and I think Torie's goal is addressed.

[muon2]

On Muon's map, there were subunits in Puget Sound, offshore Federal Way and Des Moines, but not offshore Normandy Park and Berien, because those two cities have annexed the area in the water. The same thing happens along Lake Washington.

Including these areas creates contiguity, and may foster an impression of connected, for example in this case, with Vashon Island, which should only connect based on the ferry (and the ferry landing).

In Florida, the area within the 3-mile limit is included in their maps, and within their compactness scores. It is gamed, along with the Everglades.

This one confuses me some. We don't do compactness. We require bridge or ferry connections. If we have an instance where municipal water surrounds an area not within the municipality, and that area not within the municipality has adequate connections to another CD, I am leery of using the water to merge that water surrounded area to the municipality, but on this one my views are more tentative.

Erosity is a compactness measurement.

The judge in Florida just approved a plan that used the Everglades to bypass population.

Yes, but we don't use it. It is one of several advantages of using the highway cut proxy as a measurement of erosity. It allows for benign rather than partisan erosity - erosity that is really not about people, but geographic obstacles, or county lines which themselves are erose.

How would you define subunits in Miami-Dade County?

I also think you are too narrowly tied to a particular methodology.

How would you in general define nested sets of units that districts may be composed of?

Since FL has county school districts, they obviously can't be used for subunits. My first inclination for Miami-Dade would be to use the method I used for King, but substitute Census CCDs for school districts. That is I would first make each incorporated muni in the county a subunit, then I would group the contiguous areas in each CCD as a subunit. Small isolated pockets of unincorporated areas would attach to adjacent munis in the same CCD.

[muon2]

Torie has been heavily advocating for SKEW reducing plans of late and it's an important concern. I've reviewed our MI scoring from a year ago and the thread is worth a reread (it's only three pages long). The exercise was heavily data driven with tables and some regression analysis applied to the scoring metrics.

One of the things we did was address train's strong concern that inequality matter. I share some of that concern and the solution was to add INEQUALITY to CHOP for that arm of the Pareto test. It provided that a chop was ok if it was making the inequality substantially better. We already were applying the MI rule against multiple shared chops between pairs of districts.

If the goal is to not penalize one-bite chops that lower SKEW, then the obvious solution seems to just add SKEW to CHOP. Couple that with the MI rule that no two districts can share chops with more than one subunit and I think Torie's goal is addressed.

Oh my. Ouch! First, inequality does not matter. It has next to zero public policy importance. It should only be used as a tie breaker at the end of the road, and maybe not even that. Perhaps both maps otherwise tied should be put into the eligible to pick pile.

If skew is added as a third prong to the pareto optimal test, we might as well go home. Messy gerrymandered maps with zero skew will be put into the eligible pile. Then what is the point?

I appreciate Muon2's desire for elegance, with every subunit equal in status and so forth, and every rule either a green light or a red light, e.g. bridge chops. But sometimes elegance gets in the way of the practical and the good, and that is the case here. I think my suggested language is the right balance. It will reduce skew without degrading the quality of the maps, and avoid the catastrophic scenario. Folks will be more accepting of accidental population accidents that materially affect partisan balance when it entails respecting the integrity of counties and cities, and to a lessor extent, townships. They will not when it comes to respecting the integrity of these artificial units, unless a map gets too erose.

I know that I am repeating myself. I won't change my mind absent a good reason to do so here, of if I have missed something, or whatever. Change I know is hard, but here it is really, really necessary, in my opinion.

I'm not sure you reread the thread I linked. I hope you did. I think train spoke for what many would consider good public policy goals. You do too, just different ones.

I also think you should see that the addition of inequality did not undermine the impact of the chop count, but it did build in a natural trade. Maps with near zero inequality did not emerge because the chop count would have shot up by more than the inequality came down. By the same token a messy plan with a skew of zero probably requires so many chops that it will easily lose to a good low-chop plan with a slightly higher skew. If a skew of zero doesn't require a bunch of chops, then I think you would agree that it should survive.

In any case, note how we really tabulated the impact of the factors in that MI thread. The same data-based approach would see whether your fears would be realized or not. My concern with the one-bite rule is that it will be abused for political purposes as in MI. My proposal essentially is a one-bite rule that only applies if skew improves.

[muon2]

I will read that thread now, if I can find it, or you linked it. Now you have me more confused. It does not matter who much a map with a better score on one parameter degrades another. Both maps make the cut. So a horrible map to get to a zero skew makes the cut, if there is no less horrible map out there that does.

I think I see one point of confusion. I was not suggesting SKEW as an independent Pareto variable. I was suggesting that the Pareto variable be CHOP + SKEW. That eliminates the incentive to use a one-bite that does not reduce SKEW, such as protecting a residence of an incumbent.

[muon2]

I will read that thread now, if I can find it, or you linked it. Now you have me more confused. It does not matter who much a map with a better score on one parameter degrades another. Both maps make the cut. So a horrible map to get to a zero skew makes the cut, if there is no less horrible map out there that does.

I think I see one point of confusion. I was not suggesting SKEW as an independent Pareto variable. I was suggesting that the Pareto variable be CHOP + SKEW. That eliminates the incentive to use a one-bite that does not reduce SKEW, such as protecting a residence of an incumbent.

I see. That is a somewhat less insane proposal. But that would have in the population accident scenario, potentially the lower skew map tied with the higher skew map. The lower skew map needs to win. Period. If by a series of accidents, one of them being due to a population accident that avoids any non Voting District subunit chop, one can use a chop to  accommodate an incumbent, well so what? If is very unlikely to happen, and probably there is another map with an identical score with slightly different lines, that will not. In fact there could be a ton of such maps, that vary be a precinct here or there. And no, inequality won't decide, because you can chop precincts to keep the populations identical, assuming that the two CD's in question, or one of them, was at the end of the population bell curve. Which suggests to me, that the map with the "best" PVI should win here, the best being the one that moves the PVI in the direction of the party at the short end of the skew stick, even if not into another skew category.

Am I right there that there is no penalty for chopping subunits for non-macrochops, assuming no more road cuts are in play? I asked that question before.

In the MI exercise each chop at each level counted towards the sum. We didn't allow subunit chops unless there was a macrochop, so the situation in your question couldn't arise. This was our rule then:

Raw CHOP score

Here's how I would write that into a scoring rule for MI.

Definition: Subunit. The geographic subunit of a county are the census-defined county subdivisions. Except for Detroit, the geographic subunit of county subdivisions are the vote tabulation districts (VTD). The geographic subunit of Detroit is city-defined neighborhood cluster, and the subunit of the neighborhood is the VTD.

Definition: Chop. A single chop is the division of a geographic unit between two districts. A second chop divides the unit between three districts. In general the number of chops is equal to the number of districts in that unit less one.

Definition: Macrochop. A macrochop exists in a geographic unit when the remainder of a geographic unit after subtracting the population of the largest district in the unit exceeds 5% of the quota.

The CHOP score is the sum of the chops of all counties, plus the chops of all subunits in counties with a macrochop. County subunits may not be chopped in counties that do not have a macrochop.

[muon2]

Ah, the classic Torie C. Low erosity in exchange for some chops. With a low SKEW I would think my summation proposal would appeal by making it even stronger. Why leave anything to chance?

Anyway what I want to work on is the erosity metric for the non-township states of the south and west. Is there agreement for my WA subunit creation algorithm?

For example, consider these items for the following section of King.
The unincorporated places south of Seattle are grouped together as a single subunit (pop 32,234).
The small unincorporated areas north and east of Renton are grouped with Renton (pops 494, 2100, and 2884)
Issaquah (pop 30,440) is largely but not completely surrounded by the unincorporated area of its school district (pop 27,338) , so that a district could have an unusual bump if it had one but not the other.

[muon2]

Here's northern King with more contrast.



I started with the incorporated munis, which also have nested precincts. Wholly enclosed unincorporated areas went to those munis.

Then I went to school districts using this map. WA has detailed atlases of the districts online that I used for exact boundaries.


The Northshore school district includes the cities of Bothell, Kenmore (pop 20460), and Woodinville (pop 10938). The single precincts east and west of Bothell were less than 0.5% of the quota and they were added to Bothell (total pop 19244). I could have added the area west of Bothell to Kenmore which is slightly larger, but the area is entirely in the Bothell zip code.

The area south of Bothell in the Northshore district includes 5615 residents (light green precincts), so it is its own subunit. The Cottage Lake area east of Woodinville in the Northshore district includes 20396 residents (spring green precincts), so it is its own subunit.

The Riverview school district includes Carnation and Duvall. The unincorporated area completely surrounds both those cities so they are all grouped as a single subunit with 17502 residents (orange precincts).

The Lake Washington school district includes Kirkland (pop 48787), Redmond, and the majority of Sammamish (pop 45831). The disconnected parts of Redmond are uninhabited conservation land so they are ignored for the purposes of subunits. The precincts between Kirkland and Redmond have less that 0.5% of the quota and are added to Redmond (tot pop 54510) as is the uninhabited precinct for Marymoor park on Lake Sammamish.

The Finn Hill area north of Kirkland in the Lake Washington school district has 30394 residents (orchid precincts), so it is its own subunit. The Bear Creek area east of Redmond has 20501 residents (pale lavender precincts), so it is its own subunit. The ragged spot on the northern edge of the Bear Creek subunit is due to the Meander voting district, a precinct with more population in the Cottage lake subunit.

The subunits described above range in population from 5615 to 54510. The smallest one could arguably be below the minimum, at which point it would be added to Bothell. A threshold of 0.5% of the county population would set the minimum unincorporated subunit at 9656.

[muon2]

Here's my division of central King.



The Renton school district includes the cities of Newcastle (pop 10453) and Renton. The Renton school district barely extends beyond the city limits to the north and east creating small pockets of under 3000 people. They are merged into Renton (tot pop 96459). The Bryn Mawr-Skyway unincorporated area to the west is its own subunit at 15645 (pale pink precincts).

The Issaquah school district includes the cities of Issaquah and the south half of Sammamish. There is a small enclosed area as well as a small area between Issaquah and Sammamish merged into Issaquah (tot pop 30460). The rest of the school district extends south and east to Mirrormont and it is its own subunit with 27318 (cream precincts).

The Snoqualmie Valley school district includes the small cities of North Bend and Snoqualmie. They are completely enclosed by the unincorporated area of the district so it is all gruped as a single subunit with tot pop 37156.

Note that the incorporated town of Beaux Arts Village is surrounded by Bellevue and was merged into it along with some unincorporated inclusions for tot pop 128192. Shown on the northern map the other small communities of Clyde Hill (2,984), Hunts Point (394), Medina (2,969), and Yarrow Point (1001) were not included since they are not surrounded by Bellevue. It's similar to the Sylvan Lake-Keego Harbor-Orchard Lake Village triad in Oakland county MI.

[muon2]

Here's the high-contrast detail for SW King.



The Federal Way school district has the city of Federal Way (pop 89306) and the unincorporated Lakeland area to the east. Lakeland is in two discontiguous subunits, the north part has 12976 (yellow green precincts), and the south part has 7982 (pale green precincts). If the threshold were raised from 0.5% of a CD to 0.5% of the county, the south part would be merged with Federal Way.

The Fife school district is based in Pierce county, but includes the part of the city of Milton in King (pop 831). It also includes the small unincorporated Trout Lake area with 3592 residents (gold precincts). These two would be merged with the higher population threshold.

The Auburn school district includes the cities of Algona (pop 3014), Auburn (tot pop 63076 with inclusions), and Pacific (pop 6514). The unincorporated Lake Holm area east of Auburn has 6894 residents (lilac precincts) and is a separate subunit at 0.5% of the quota, but would combine with Auburn using a 0.5% of the county standard.

The Enumclaw school district includes the cities of Black Diamond (4151) and Enumclaw. The unincorporated part of the school district completely surrounds Enumclaw but not Black Diamond, so  Enumclaw is merged with the unincorporated area for a total of 23296 residents (red-orange precincts).

The Tahoma school district includes the city of Maple Valley (pop 22753 with inclusions). The rest is the unincorporated Hobart area which forms a subunit with a pop of 14819 (light blue precincts).

The Kent school district includes the cities of Covington (pop 17575) and Kent (tot pop 93012 with inclusions). The unincorporated Berrydale area south of Covington with a pop of 9,981 (beige precincts), and Lake Youngs area north east of Kent with a pop of 48,462 (tan precincts) are separate subunits.

Not entirely on the map is the Highline school district with the cities of Burien (pop 33313), Des Moines (pop 29673), Normandy Park (pop 6335), and Seatac (26909). More significantly the North Highline area between Burien and Seattle is all unincorporated with some separate marked places, but it forms one subunit with a pop of 32254.

[muon2]

Here's the high-contrast detail for SW King.



The Eunumclaw school district includes the cities of Black Diamond (4151) and Eunumclaw. The unincorporated part of the school district completely surrounds Eunumclaw but not Black Diamond, so  Eunumclaw is merged with the unincorporated area for a total of 23296 residents (red-orange precincts).

Enumclaw (note spelling) has some exclaves, including one that straddles the Pierce-King line. Does that mean the school district doesn't enclose Enumclaw.

Thank for the spell check. I've fixed it in the main post. The exclave on the county line is a park with no people. I ignored it for connection purposes as I did the exclaves of Redmond.

Beaux Arts Village is contiguous to Mercer Island, but not connected due to Lake Washington. Since the only connection is to Bellevue, it is surrounded.

[muon2]

Beaux Arts Village is contiguous to Mercer Island, but not connected due to Lake Washington. Since the only connection is to Bellevue, it is surrounded.

Is this different than Asotin or San Juan counties?

If Bellevue was chopped, could I dice Beaux Arts Village as part of the chop?

When I'm shifting counties to form regions I treat Asotin and Garfield as a single unit. Any shift of Garfield brings Asotin along with it, since Asotin is only connected to Garfield. A bigger example is the UP of MI which stays together with Cheboygan for CDs since it has only that one connection.

Bellevue is large enough to macrochop so it's conceivable that neighborhoods could be defined, and Beaux Arts Village would be a neighborhood. If we treat Bellevue as we did other large cities in MI (except Detroit), then the subunits of the city are the precincts and Beaux Arts Village is its own precinct, so it would not be chopped.

[muon2]

The subunits of a city are its precincts? What does that mean? That seems to imply there would never be a chop. When does a city have neighborhoods larger than single precincts?

I was quoting the rule we adopted for MI last year. We decided that we would only define neighborhoods for Detroit, since it had to be chopped. No one raised the issue of neighborhoods for the other communities, so we defined precincts as the subunits for computing erosity had the situation occurred. It's in the thread I linked.

The MI rules implied that cities that must be chopped would have defined neighborhoods, but we never addressed what if any other cities would have neighborhoods.

[muon2]

Torie, what are your criteria for where you did split swaths of unincorporated area into separate subunits? We should be consistent. I still think it makes no sense to absorb included unincorporated areas into cities, but not minimal pockets trapped between two cities.

I don't have a problem if we want to make all the incorporated munis stand out even when surrounded. I don't think it will make a difference in the end, but that's just my opinion. I would note that towns are recognized as incorporated units. There are four in the county - Beaux Arts Village, Hunts Point, Skykomish, Yarrow Point. I would think that you want to treat these towns the same as cities.

[muon2]

Your criteria strike me as an artist's endeavor and not something one could codify. I tried to be quite explicit that the divisions in the unincorporated areas should be based on the divisions between school districts. You haven't given a reason as to why that is unacceptable.

One way to deal with pockets is if the school district approach is used then movement of separate pockets or inclusions count without chop penalty. We would treat them as we would a disconnected county or muni (or township as in OH).

My approach to this assumes the MI rules as a baseline, but not necessarily the endpoint. The MI rules have the advantage of more than a dozen good map attempts from our different mappers to provide a rich set of data to test the rules. Given that baseline, I want to see where individual rules in the baseline fail before making changes. My hope is that the WA exercise will allow mappers to craft examples that show the weakness in those rules, but we aren't there yet until we can agree on subunits that replace the MI townships.

Along that line, I think it is unwise to try to encode neighborhoods for all large munis. Given the lack of agreement on county subunits, I imagine it would be that much harder to come to agreement with the Bellevues of the state. Munis and school districts are encoded Census info, but neighborhoods are not. I propose that neighborhood definitions are only required for those cities that must be chopped based on population or for those where the VRA compels that a city be chopped.

[muon2]

Your criteria strike me as an artist's endeavor and not something one could codify. I tried to quite explicit that the divisions in the unincorporated areas should be based on the divisions between school districts. You haven't given a reason as to why that is unacceptable.

It has nothing to do with artistry really (I understand now that you are after me any time that you use the word "art" or "artist").  It can all be codified. In this case, it is unincorporated territory trapped between cities, that is not of sufficient size to justify creating a separate subunit for such territory, as opposed to serving as a vehicle to encourage line chops there. I gave the policy reason. To encourage the chop there, rather than in the actual city itself. It seems rather obvious to me, and has no policy downside - just upside. I am using your school districts for unincorporated territory, that is not of small size, trapped between cities.

I raised my point because sometimes you use the school districts to divide unincorporated areas but other times you do not. I think the sounder policy is to use it in all cases - it's much easier defend that way. If it creates some small pocket pinned against a city, such as I have east of Renton, so be it. It is no different than the issues faced in OH where incorporated munis create all sorts of fragments out of the townships. Stark county is a good example.



Why not view King like Stark? Start with the school districts instead of the townships, then overlay the munis on top as whole entities. There will be lots of fragments in King, but like in OH, we won't penalize subunit chops that act on disconnected fragments.

[muon2]

At one time we held that if township consisted of two disconnected parts and the parts were placed in separate districts, no chop penalty was assessed. Did we change this at some point? Should we?

WA has no township structure. The Census encodes blocks by munis, townships, and school districts where any of these exist, so they are all defensible redistricting subdivisions. Residents in areas with townships are far more likely to know their school district than their township (trust me on this one). Therefore school districts are a reasonable substitute for townships in states like WA that have no townships.

[muon2]

At one time we held that if township consisted of two disconnected parts and the parts were placed in separate districts, no chop penalty was assessed. Did we change this at some point? Should we?

Don't know. We probably should. It should be avoided where possible. Perhaps it should be another preference item. It is better to chop that township, than it is another one, without such a municipal barrier severing it. But rather than take in another township, it would be better to take in the other severed portion of the township first.

WA has no township structure. The Census encodes blocks by munis, townships, and school districts where any of these exist, so they are all defensible redistricting subdivisions. Residents in areas with townships are far more likely to know their school district than their township (trust me on this one). Therefore school districts are a reasonable substitute for townships in states like WA that have no townships.

I think I have made it plain, that I agree with that. If you are saying that folks would rather that  a city or town be chopped, rather than a school district, I tend to doubt that. Townships (NY calls them towns, because they are weird), are very important to folks in Columbia County, NY, I assure you. Granted, in Madison County, Iowa they are not. I don't think they have any governmental structure. They are just there on the map, without any meaning. In NY, patronage is associated with towns/townships, and jobs and fighting over money, and of course, town judges, and town council members, and who gets elected a supervisor from the town, and so forth. But then, New York loves lots of layers of government. It means more patronage! And Hudson has 5 county supervisors (currently elected from illegally shaped districts until I sue to put an end to that, but I digress), while each town has but one. That means 4 extra jobs that pay about 20K a year for very part time work, but health benefits. Since industry is dead largely in upstate NY, for the locals, who don't work over the internet, or bring their money with them, it's a big, big deal. It's either a government job, or working at Walmart, or being on welfare.

Sorry for the NY diversion. NY fascinates me. It's another planet from the one from which I originated.

Of course most people know their city and state ahead of anything else. IIRC in NY, New England, and a couple of other northeastern states munis can't spill over town lines so towns have an elevated status. In New England towns are so elevated I would ignore counties entirely for the rules (except for those counties with unorganized territory) and just use towns. Experience with towns there doesn't transfer easily to other states outside the Northeast.

I was thinking of the Midwest when I made my comparison between townships and school districts.

Throughout the Midwest, there are townships. Most have an actual government structure with elected trustees. IA does, too. If I grab random people from IL, IA, OH, etc, they are more likely to know what school district they are in than what township serves them. That's especially true in urbanized counties near the big cities.

That's why I think it makes sense for counties like King in WA to use school districts where we would use townships in counties like Stark in OH. The rules then become easy to defend and easy to generalize to many states.

[muon2]

What exactly are we disagreeing about now? Are we still on the trapped unincorporated thing, where it's a choice between chopping a city or town, or a school district? Or have we moved on to something else? Or is your order city, then school district, and then town in WA, because towns don't mean anything in Washington, or what?

My order is city (including towns like Beaux Arts Village which do have meaning) then school district. There is no other division at the county subunit level.

My little white spots are just between cities anyway. You just want to have cities swallow up these little bits, so that then when chopped, no distinction is drawn about where you chop. Oh, and I think you have some cities swallowing up non trapped areas, rather than keeping them in a separate subunit, to exacerbate the matter. This concept of cities swallowing up stuff is novel. We have no precedent for such subunit expansion.

I have already agreed to abandon that. I have replaced it with the model we use in OH but using school districts instead of townships. It's a model for which OH gives precedent.

What I thought we were dealing with, is how to handle territory that was bereft of subunits. But then we had mission creep, and creeping municipal lines effectively. What you really want to do, is dump cities and towns into the dumpster, and just have school districts really, simply because we have some territory that is unclaimed, so let's deal with that, by having wholesale erasures. Well maybe.

Cities and towns remain supreme as county subunits (just not Census places - they don't exist in WA statute). The only question is what to do with the rest.

We are not making much progress are we?  

We would if you'd let me concede the points that I have, and then consider alternatives. What I won't do is put expectations of what SKEW will do ahead of defining subunits. It must be a completely independent exercise.

[muon2]

Can you change your green color to red or blue or something? I can't read it. Sorry.

ok

[muon2]

Can you change your green color to red or blue or something? I can't read it. Sorry.

ok

Thanks. OK. What exactly are the open issues then, being specific as you can, with Torie wants that, and Muon2 wants the other, and the reason is this?

Skew is everything. This needs to be reversed engineered, so that we are confident that we can say, we have done what we can to reduce skew consistent with good maps, the types judges would like. As I have said before, that mostly turns on what is done within big subunits, not without. Thus the one bite rule, consistent with erosity patrol. That way, if the skew is not reduced within big subunits, that is because to do so, would tend to make the map ugly, and judges will not like that. What is done without, is mostly a matter of common sense, and practicality, and nothing else. Not much really turns on that as a partisan matter, absent a population accident, maybe, and maybe never, on a systematic basis.

You just laid out our greatest point of disagreement. There are too many experts on both political sides as well as among good government groups that will be highly suspicious of a system designed for a political outcome, including one so laudable as political equity. Some good gov and third party advocates would say a SKEW of 0 with a high POLARIZATION is the worst kind of bipartisan gerrymander where both parties lock in their seats for a decade. Some groups would trade SKEW for low POLARIZATION so that the candidates would be more responsive to the electorate. Look at the number of states that have said political data should not be part of the redistricting process at all.

The system is about balance of all the variables: SKEW, POLARIZATION, INEQUALITY, CHOP, and EROSITY. The Pareto set starts with the balance of CHOP and EROSITY. As you noted in your other recent post states can modify the CHOP with items like UCCs, MCCs, as well as the other variables to suit their priorities. A lot of effort went into our MI work to test the balance of the variables. When one of the variables gets built into the design of the other variables, then the Pareto balance starts to lose its meaning other than as a filter for that primary variable.

[muon2]

Torie, what are your criteria for where you did split swaths of unincorporated area into separate subunits? We should be consistent. I still think it makes no sense to absorb included unincorporated areas into cities, but not minimal pockets trapped between two cities.

I don't have a problem if we want to make all the incorporated munis stand out even when surrounded. I don't think it will make a difference in the end, but that's just my opinion. I would note that towns are recognized as incorporated units. There are four in the county - Beaux Arts Village, Hunts Point, Skykomish, Yarrow Point. I would think that you want to treat these towns the same as cities.

I think that we need a clear statement of principles, that individual States can adapt to policy appropriate to their circumstances.

At best we are doing is experimenting with policies that Washington might adopt.

Here is a statement of policy on county subunits.

Counties in each state should be divided into subunits. County subunits should cover the entire county, though counties with less than 10% of the quota may be covered by a single subunit. (The rationale is that such counties cannot be macrochopped.)

County subunits should be based on political subdivisions recognized by the Census. These may include entities like cities, towns, townships, reservations, and school districts. CCDs defined by the census may be used when there are insufficient political entities to cover a county. When possible all counties in a state should use the same method to determine county subunits.

County subunits may be based on more than one type of political subdivision. For example a county may be divided into cities and townships. In the case of overlapping political units, there should be a clear priority as to the assignment of a parcel. In the example the priority may be that a parcel in both a city and township is assigned to a city.

County subunits may have discontiguous parts. In those cases there is no chop counted when the subunit is split such that discontiguous parts are kept whole. Discontiguous parts may contribute to erosity based on the nature of a split.

Here's how King would look using the preceding statement with cities and towns as the primary and school districts as the secondary basis for subunits. There are 55 subunits in all: 35 cities, 4 towns, and 16 school districts, many of which have multiple fragments. VTDs (precincts) are used to approximate the actual school boundaries.

[muon2]

I don't think I disagree with anything stated above. In part, that is because it is generalized. The devil is in the details. I was asking for where we disagree on the details. I laid out my point of view on these matters. Of course skew does not rule, nor does anything else. Everything is a balancing test. I came up with the one bite rule working with Phoenix. That is consistent with good maps (given the erosity constraint). It is probably what a judge would do. I would certainly do it as a judge. And I would not like a population accident to foreclose what should be done. I see I think on your map those little white bits there that were on my map, now colored in. Does that mean they can go in either adjacent CD? Is that the Stark County issue, where you have township subunits divided by a city, and I suggested a preference rule?  If so, how do our maps differ, other than that you appropriately added the towns?

As I said, what I am looking for now is a specific list of disagreements on the details. It seems when it comes to what the subunits are (as opposed for the units as to how to manipulate them into CD's), at least in King County, that maybe we don't have any disagreements, anymore. You take territory that is not in a city or town subunit, and put in another kind of subunit that is workable. School districts for such unassigned territory seems to be workable in King. School districts do not erase the lines of other subunits, even if they overlap them. So they are fragments of school districts in essence, in many cases at least.

I don't want to agree to anything anymore, until I fully understand it, and its implications. I don't want the estoppel thing to come up, ever again, if possible. I have been beaten up enough on that one already.

One detailed difference is with areas like the one between Bothell and Kirkland. I have it split between two school districts. The few precincts just south of Bothell are like the SE corner of Plain in Stark. They are adjacent to a larger unincorporated township, but they are kept separate as a fragment of Plain. I do the same in King. There a number of similar places where you have merged small unincorporated areas that are in separate school districts.

A second potential difference is how those fragments are handled. Yes, a whole fragment can be shifted to a different district with no chop penalty - that's the Stark equivalence. However, it's still a chop and that means the shifted fragment could contribute to erosity, just like a chop anywhere else. I say could, because it could work either way or have no effect, again that's just like what happens with how we've applied erosity to chops in other states. To make it clear I show all fragments in the same school district with the same color.

I think my resistance to the one bite rule stems from how well our maps have performed without such a rule. Your MI offering is an excellent example that worked without the one-bite. I am resistant to disrupting that kind of success. I'd like to see that a particular goal is impossible without one-bite.

I'm also resistant to an endless preference list that isn't part of the score. I think one of the successes of the UCC model is that we made into a scoring modifier, not a mere preference. There may well be other items that should function like the UCCs and that means working out a scoring rubric for them. My example along this line is the MCC to handle the rural minority counties of the South. We should look at each preference and see how to create a balanced score for it.

[muon2]

This is my proposed subunit map for King County.



It was generated directly from block-level census data, which includes. among other data, for each census block: (1) census place (city or CDP); (2) school district; (3) population; (4) land area; and (5) water area.

It was used in conjunction with place data (for King County); and school district (for Washington).

Using Excel, a count of census blocks for each school district was obtained. This isolated the list of school districts in King County, either wholly or part. We treat a city or school district that spans the border as being separate entities.

Next using a two dimensional array of place X school district, the population of the intersection of place and school district was calculated.

For each incorporated city, its largest school district was compared to see if it exceeded 80% of the total population of the city. If it did, then the blocks in the city were reassigned to the predominate school district. (the original school district is retained. A new column was calculated with the assignment to the predominate school district, now a proto-subunit).

For the other three cities, which are divided among districts, (Black Diamond, Sammamish, and Newcastle), blocks were assigned to city-specific subunits. These subunit/cities are shown in red.

Finally, blocks that satisfied all of these conditions: (1) no population; (2) no land area; and more than a km² of water area were assigned to a special subunit. This did not work particularly well. It picked up inland waters and missed areas of Puget Sound. There are feature files that associate specific blocks with water features, which could be used for Puget Sound and other external waters.

The block shapefiles were loaded into QGIS, and joined with a .csv file that contained the Block GeoID-subunit pairs. A new layer was created by dissolving the boundary between blocks in the same subunit. That is what is shown here.

The dissolution was not particularly fast (hours rather than minutes or seconds), but King County has 35,838 blocks.

I like the map, and it seems to also match my principles with a different prioritization. My concern returns to the nature of incorporation in WA.

The city of Lake Forest Park was incorporated specifically to have an identity separate from the larger Shoreline school district. That's why I like the OH model better here, communities that incorporate get a higher level of recognition as subunits. I think my map creates that same type of recognition for incorporated communities.

[muon2]

I saw the UGAs and PAAs on the King website. I prefer the subunit principle that requires county subunits use boundaries that are recognized by the Census. It's easier to generalize to other counties and states, and the source of the definition is clear.

[muon2]

I saw the UGAs and PAAs on the King website. I prefer the subunit principle that requires county subunits use boundaries that are recognized by the Census. It's easier to generalize to other counties and states, and the source of the definition is clear.

The census maps are out of date. I think the aggressive annexation policy started in 2009.

I would expect that the units would be defined around 2018. I think the census geographic data is released in December of 2020, so that the redistricting, etc. can be programmed in anticipation of the PL 94-171 release in the spring. Incidentally, cities in Washington are required to redistrict city council districts within NN(60?) days of receipt of the data from the Washington Redistricting Commission. It appears that they just do prep work, rather than actually get involved.

AFIAC, so long as the unit boundaries conform to census blocks, they are fine.

I would think that a state would be more likely to adopt your process if they have a role in defining its use.

I'm basing these maps on 2010 as a test. There's no question the subunits would be different going into 2020 because of annexation. OH has a similar effect as incorporations and annexations cut into townships, leading to different subunits n 2020 than in 2010.

The preferred role for the state is to pick a priority among subunit choices that would apply to all counties. The recommended subunit choices would be those that would be coded for blocks in the decennial Census. If WA creates county subdivisions for use in the 2020 census, then those would naturally be the best choice.