Local vs regional road connections

[jimrtex]

The network map I posted was in my Atlas gallery that I found through a search. I created it in 2012 using Visio. It was for a thread about connections and regions in WA that had a lot of back and forth between jimrtex and I. Our differences about connections in NC can be visited in that thread three years ago. We really keep crossing the same ground on these issues, but I tend to resist wholesale thread necromancy so here we are.

Here is my Washington network map.  The fine red lines indicate adjacent counties, that I classified as unconnected. As usual, there are differences, between my definitions and Muon2's.



This is a version that might be presented to the public as part of a simple app for creating county combinations. The links are shown as traversing the boundary, and boundaries between adjacent but non-connected counties are shown as a barriers.



This is a district map with no chops.



This divides King County into 3 districts. IIRC, I used school district boundaries for the secondary division.



This adds a split of Pierce County to reduce deviation to a reasonable level.



This is a simplified county outline map showing multi-county UCC's and connected/unconnected status.

[muon2]

The problem with muni chops in WA is similar to what we encountered in MD, NC, FL and other states that lack subcounty government. Some of the places in WA on DRA are incorporated places, but others are not. I would not want to give particular importance to unincorporated place lines, since they don't have to reflect real boundaries. We could use the Census CCD boundaries in WA counties, much like using the MD election districts. However, in King they are mostly too large on the west side, so they would need to subdivided based on incorporated munis. Defining the subdivisions in King and Pierce are a necessity to look at connections and erosity.

Yes, I agree with all of that, and am inclined at this point, absent something persuasive, to follow the precinct nest rule. But in individual states there might be a good reason not to do that, and the state should just change their precinct lines. But the fact that the precincts are not nested, suggests the states with that syndrome don't care much about the integrity of such localities. The same might obtain for city neighborhoods. It would be a lot to ask for say NYC to have to change its precincts wholesale to accommodate your idea of what and where the neighborhoods are. But if the precinct lines follow the neighborhood lines, great. Bremerton is an incorporated city. You just missed its erosity on the south end. No problem. It's just that you so rarely make mistakes, that I tend to think what you do is deliberate. In this case, I thought it might be a pavement obsession.

This is the issue most germane to the thread. WA generally has precincts nested in cities and towns but not Census places or incorporated towns (all of which have the same marking on DRA). Precincts for cities seem to have either a name or prefix number to indicate that they are from that muni, but I've found some exceptions.

The most important property of county subunits is that they must cover the whole county. That means the all the non-city areas have to be accounted for.  My sense is that we are using cities where there are nested precincts and single precincts everywhere else. That's a lot of county subunits, but one can usually extract it from DRA. Using them as subunits for erosity will tend to make higher scores in the areas defined by individual precincts.

jimrtex suggested school districts which in King look like this.

They aren't so numerous that they would run up erosities, the public has a strong sense of being in a particular school district, and they aren't going to change at the whim of the county board. However, it doesn't look like they line up with the DRA precincts. The best we could do is make an approximation of precincts to school districts.

What works best from your perspective?

[Torie]

How do the the school district boundaries interplay with the boundaries of towns and cities? Are towns and cities nested in school districts? If not, using school districts will not work. We don't want to start deliberately chopping cities and towns. If they are nested, then I think Goldilocks needs to be made happy. If you have subunits that are too big in population, or potentially in geographic area if they start cutting off stuff that a CD needs to go through to have pavement connections, or whatever, you start getting chops, which is a scoring disaster, and we already have rules that make one avoid macrochops absent a compelling reason to do so. We don't need any further disincentives there. If they are too small, you then have a zillion road cuts, wrecking the erosity score. Using individual precincts for anything is just silly.

Ideally, one should use towns (townships) and cities and ignore villages (other than to have a rule to limit the number of village chops where one has a choice where the lines go), the way we do in much of the Midwest and Northeast. If there are areas in a county that have neither, and that area is too big in population (if relatively small in population, it really does not make any difference, so draw the lines where one wants, being sensitive to pavement connections ala say our Yakima County macrochop), then for those areas coming up with something else makes sense if available, and it does not interfere with towns and cities. So in King, if there is no nest problem vis  a vis towns and cities (I assume that the precincts are nested in towns), using school districts for the east side of King that does not seem to have any subdivisions (or almost none) would make sense because Washington does not have township boundaries that take in all of its real estate. On the west side, where all the real estate (or almost all of it) is defined by towns and cities where the precincts seem to nest), I would ignore school districts. We don't want a situation where you get one chop penalty for chopping a school district, and a second one for chopping a city or town at the same time within the school district. It's all a balancing test.

I would note that in order to put Vashon in the Seattle CD, we both chopped that school district north of Seattle in our maps by severing off the town or city on its east side. So we want to ignore that school district, and my ramblings above provide a rationale to do so.

Playing around with the mapping tool around Carnation, I would note that the precincts do not nest with the school district line around there. And I really don't think the "approximation" test is practicable. So to make it work, Washington (or King at least) would have to move its precinct lines. And unlike MD, with its voting districts, there seems to be no workable substitute. So absent changing the precinct lines, I think we are just back to avoiding town chops were towns exist (like Carnation), and other than that, one can do what one wants with the stricture of trying to have state highway connections if possible to do so, and if not, then at least pavement connections, where that is possible with a certain line selection that does not precipitate chops. It's hardly a disaster in the sense, that on the east side of King, the population to play with is really not that large, nor the political affiliations from one area to the other all that different.

[muon2]

To guide our thinking, I put together a map of King with the incorporated munis from 2010 colored in. This image shows most of eastern King including the place lines from DRA. There are over 320K people in the unincorporated parts, so they need some reasonable assignment to function as county subunits.



It's clear to me that any hope of precinct nesting will fail for anything other than incorporated munis. In order to test the metric I need subunits for erosity. Therefore to avoid a profusion of precinct sized subunits, there must be precinct aggregation, even if it is only approximate.

Muni growth in King is driven by annexation, and the county actively encourages built-up areas to annex to an adjacent town. The King web site includes descriptions of potential annexation areas that it is promoting. Since 2010 the area north of Kirkland got pulled into Kirkland doubling its pop. There are a half dozen current PAAs (potential annexation areas).

The area east of Renton is a PAA that rejected annexation in a 2007 referendum, but the county is still pushing it to get greater efficiency of services. Note that a bit of the PAA has already been pulled into Renton.



Are these useful for creating county subunits?

[Torie]

The areas of potential annexation being appended to the city that might absorb them, or their own locality, or what? Do these potential annexation areas have nested precincts? If not, given the relatively small population involved as compared to what is adjacent, why is this really important? Not that it matters for rule making (and I strongly agree on trying to have a grand unified theory across the Fruited Plain, if only for the "elegance" of it all, itself an object d'art, just like my desire for straight clean lined rectangular CD districts), but given the partisan geography, it will not make any partisan difference in any event in WA. I wonder in how many states, is this conundrum in play, that really matters? By really mattering, I mean real estate in a county without nested precincts in something we can use, if only because it is there, if not really associated with anything else at all that matters much, ala MD, where exactly how the lines are drawn, might really matter? If it does not really matter, we need not worry about gratuitous erosity, I would think.

Oh, one other thing. Assuming the hinterlands of these highly populated counties have no sub-jurisdicitons, that are precinct nested, and where the exact lines are drawn might make a difference, isn't this a good venue to try to get the SKEW down? Just saying ... . I bring this up, because I just did that with Maricopa County based on 2020 census projections. It would be unfortunate, if some neighborhood thing prevented that. Unlikely, because my lines were clean, and whatever neighborhoods in Phoenix are created, would, and should, not make much of a difference. But it might have.

I am amazed you are not busy wrapping presents or something. Have a good Xmas. The best thing about this wonderful site, is that I met you. It's been a pleasure. Best.

[muon2]

I see you continue to think about this. So do I.

If we can't find a complete set of subunits for a county then erosity can work in one of two ways. either the chops are treated simply, as we do without a macrochop, or they have additional connections to adjacent counties. Either way the only connections within the county are those from the whole of an adjacent CD. Let me use my map for King and its environs to illustrate. The primary connections are the ones that would appear for a simple chop. The secondary connections are those due to state highway crossings other than the shortest route between county seats.



CD 7 connections 4/5
Primary:
King-Kitsap (6) via Southworth-Fauntleroy ferry
King-Snohomish (1) via I-5
King-King (8 ) internal
King-King (10) internal
Secondary:
King-Kitsap (2) via Seattle-Bremerton ferry

CD 8 connections 6/7
Primary:
Snohomish-Snohomish (1) internal
King-King (7) internal
King-King (10) internal
Kittitas-Chelan (4)
Kittitas-Grant (4)
Kittitas-Yakima (4)
Secondary:
King-Snohomish (1) via I-405

CD 10 connections 3/3
Primary:
King-King (7) internal
King-King (8 ) internal
King-Pierce (9) via I-5
Secondary: none

Excluding the double counted connections between the three CDs the total erosity is 10/12. CD 8 and 10 are both D+8.

So consider this Pub attempt to give themselves a shot at a CD. All the munis are kept intact and things are connected. CD 10 grabs all the Dem suburbs next to Seattle and becomes D+13. CD 8 becomes D+4 and the Pubs carried it in 2010 for US Senate 51-49.



CD 10 looks very erose to my eye, but let's check the score without the benefit of subunits.

CD 7 connections 2/3
Primary:
King-Kitsap (6) via Southworth-Fauntleroy ferry
King-King (10) internal
Secondary:
King-Kitsap (2) via Seattle-Bremerton ferry

CD 8 connections 6/7
Primary:
Snohomish-Snohomish (1) internal
King-King (10) internal
King-Pierce (9) via I-5
Kittitas-Chelan (4)
Kittitas-Grant (4)
Kittitas-Yakima (4)
Secondary:
King-Snohomish (1) via I-405

CD 10 connections 3/3
Primary:
King-Snohomish (1) via I-5
King-King (7) internal
King-King (8 ) internal
Secondary: none

Excluding the double counted connections between two of the three CDs (7 and 8 aren't connected) the total erosity is 9/11. This plan is better on regional erosity than the first plan regardless of the method used. There's nothing in this counting to penalize the gross shape of CD 10. Yet it does matter politically. This is why macrochops need subunits - to judge the shape of urban districts nested in a big county.

[Torie]

My first problem here before we get going further, is that I don't understand your road cut count. Talk about complexity! It would seem to me the cuts generated between counties would tend to be the same for any map, where in the adjacent county another CD exists. So we are left with cuts internal to King County. In playing the game myself to minimize highway cuts, I drew your WA-10 much the way you did. Assuming non municipalities were deemed but one jurisdiction in King, we have a highway cut between Renton and Newcastle, Renton and the unincorporated area,  Maple Valley and the unincorporated area, and then road cuts however counted where the unincorporated area is chopped, which is any pavement correct? That would be a nightmare to count potentially. And you think tie breaking preference rules are complex and hard to apply (which even someone as obtuse as I think are quite simple and straight forward)!

Anyway, the incentive would be to have the CD line follow the Cedar River to minimize such cuts I suppose, and adding a road cut for Black Diamond, per the map below. Although you might say Black Diamond is not connected by a state highway so that is off the table, but I don't agree with that rule either with respect to internal cuts in a county. Any pavement will do. Anyway, it seems the the road cuts internally are quite low.

Of course, with my bridge chop definition and rule, the posited Pub temptation that is alleged to be doable with a winning score would be off the table, but I digress. It appears that we may permanently part ways on the bridge chop issue. Pity.

[muon2]

But I will get into that dealing with your King County map on that thread. On that map, did you count non state highway road cuts between municipalities in separate CD's that do not involve another county? I ask, because in general, the county chops are incentivized to go where fewer jurisdictions exist, which would tend to help the Dems because it is less populated areas that are chopped. In general, if one CD takes in the more rural areas, and the other CD takes in the populated areas, you will have one CD appending inside a county a ton of localities, with a lot of road cuts. Each and every road that leads out from a municipality to the balance of a county, even if there are no sub jurisdictions in the balance of the county, would count as a road cut. So that issue needs to be resolved first.

I'll reply over here.

The method I used to assess erosity in the King example was the one that was initially developed and applied in our exercise in AL and other states back in 2013. That only counted at most one connection between pieces of different districts within a county. We thought about applying townships in all cases, but an initial look at minor chops in OH showed that counting all the connections between townships/munis in a chopped county overly penalized those minor chops. However, something was needed to deal with dense counties like Cuyahoga where erosity could only be measured in terms of adjacent munis.

When we came to MI we dove into detail on when those connections between munis should apply. The preferred metric was to define a macrochop then apply the subunit connections only in those counties subject to a macrochop.

The essence of this erosity measure relies on having every parcel assigned to a division. For example every parcel in a state is in a county or equivalent so erosity at the county level makes sense. In MI every parcel in a county is in a township or muni so again the metric works appropriately. What we are trying to do is accommodate macrochopped counties in states where not every parcel is in a township/muni.

A big complication is that you only want subunits that reflect preexisting precinct lines. However, as I look at various state statutes I see it's usually the opposite from the direction that would work best for you. The precincts are updated to conform with political units after redistricting, not before. The policy reason is to reflect any new chops with new precinct lines.

In King there are munis but not every parcel is in one. I did not consider muni-muni erosity because I don't have a metric to define a connection when there is an unincorporated gap. For example there is an unincorporated gap between Kirkland and Bothell on my map. The gap was small enough that today they are contiguous, though they weren't in 2010. Either way, a meaningful erosity needs to assign that 2010 unincorporated area to a subunit.

[muon2]

My first problem here before we get going further, is that I don't understand your road cut count. Talk about complexity! It would seem to me the cuts generated between counties would tend to be the same for any map, where in the adjacent county another CD exists. So we are left with cuts internal to King County. In playing the game myself to minimize highway cuts, I drew your WA-10 much the way you did. Assuming non municipalities were deemed but one jurisdiction in King, we have a highway cut between Renton and Newcastle, Renton and the unincorporated area,  Maple Valley and the unincorporated area, and then road cuts however counted where the unincorporated area is chopped, which is any pavement correct? That would be a nightmare to count potentially. And you think tie breaking preference rules are complex and hard to apply (which even someone as obtuse as I think are quite simple and straight forward)!

Anyway, the incentive would be to have the CD line follow the Cedar River to minimize such cuts I suppose, and adding a road cut for Black Diamond, per the map below. Although you might say Black Diamond is not connected by a state highway so that is off the table, but I don't agree with that rule either with respect to internal cuts in a county. Any pavement will do. Anyway, it seems the the road cuts internally are quite low.

Of course, with my bridge chop definition and rule, the posited Pub temptation that is alleged to be doable with a winning score would be off the table, but I digress. It appears that we may permanently part ways on the bridge chop issue. Pity.

Actually I thought this road cut count was as simple as could be. A regional connection exists between counties where there is a numbered state highway. Within a county there is a single connection between each locally connected fragment. In a macrochopped county there is a connection between counties and fragments where a regional connection exists (used for the second number). One then proceeds to count the cuts as I did above (number after the slash).

Is there an instance in my description above where this didn't work?

[Torie]

I understand all of the above, but I don't think it is responsive to my post. Of course, we are only talking only about macro chopped counties. I accept that the precinct nesting issue should not be the tail that wags the dog. It is just an indication of whether or not there are subunits that anybody cares about for election purposes. The issue is where there are no subunits, how necessary is it really to artificially generate them?  You say one reason is that you might end up with wide partisan swings. I reply that maybe that is not so bad, if it gets the SKEW down, provided the maps don't get too erose looking. You offered up a map that looked erose, but much of that is really due to the shape around the Newcastle area, with a highway squeezing by to keep the CD connected. I grant it looks kind of ugly however. I put up a map that was not so ugly, and also did not impact SKEW, that would seem to me to have fewer highway cuts. You assert that there are more. I don't understand how. I am not sure how road cuts are counted between subunits in macro chopped counties. I thought it was any road. Granted if there is one big swath of land in King without subunits it gets to be cumbersome to count every road, where the unincorporated subunit is chopped. The Cedar River however, happens to make it all manageable in this case it would seem, particularly if any pavement connections will do within a county.

Speaking about chops of subunits in macro chopped counties, I am thinking of allowing one chop without penalty of a subunit to get the SKEW down, subject of course to a road cut count, so it does not get too ugly. The same thing within a big city. It if has neighborhoods, allow one chop to get the SKEW down, subject to the erosity rules. You won't like that one either probably. And what do you do with cities that have no neighborhoods, like seemingly Tucson, that are usable? What about Phoenix? Just create some out of whole cloth? What I do is try to have straight lines preferably following highways. Generally, in cities without apparent usable neighborhoods my chops within cities are highly "artistic" and pleasing to the eye. I like it that way. Anyway, the point being, that if there are going to be subunits everywhere in counties subject to macrochops, generated somehow, the one bite hope rule might be a workaround to deal with SKEW issues. That might be the right compromise, the Golden Mean, to resolve this particular food fight. Dogs get one bite with a negligence rap, and maps similarly get one bite in this context.

Getting back to SKEW, and Phoenix, the problem with not allowing one cut of a neighborhood, however constructed, artificially or not, is that it all becomes a population accident as to where the lines are drawn. There is next to no flexibility. It is all an accident. The only flexibility is where it happens that there has to be one chop per the population numbers. Otherwise, you are just screwed. Heck, it might end up with something that looks really erose to avoid a chop. The regime is really intolerable. It does not make a difference that matters when chopping Detroit, but it clearly might in Phoenix. It might matter as to the partisan complexion of the Statin Island based CD in NYC. It might in Tucson. I can't think of any other cities where it might matter, other than perhaps down in Texas, and then of course in some big counties. Some cities where there might otherwise be a partisan difference are driven by the VRA.

[muon2]

By my metric there are either agreed subunits that take in every parcel or not.

A) If there are agreed subunits then the usual erosity rules apply to those subunits within a county and across county lines.

B) If there are no agreed subunits then chops are assesed for munis in the county. The pieces of each district become the subunits and erosity is assessed on those (as we do with simple chops). Since every parcel is in a district the district pieces can serve as subunits for erosity.

My analysis assumed no agreed subunits but agreed munis as in case B.

[Torie]

By my metric there are either agreed subunits that take in every parcel or not.

A) If there are agreed subunits then the usual erosity rules apply to those subunits within a county and across county lines.

B) If there are no agreed subunits then chops are assesed for munis in the county. The pieces of each district become the subunits and erosity is assessed on those (as we do with simple chops). Since every parcel is in a district the district pieces can serve as subunits for erosity.

My analysis assumed no agreed subunits but agreed munis as in case B.

What does "pieces of each district" mean? I look forward to your comments as to the balance of what I said, assuming you want to keep this discussion going. Maybe you have had enough.

[Torie]

My first problem here before we get going further, is that I don't understand your road cut count. Talk about complexity! It would seem to me the cuts generated between counties would tend to be the same for any map, where in the adjacent county another CD exists. So we are left with cuts internal to King County. In playing the game myself to minimize highway cuts, I drew your WA-10 much the way you did. Assuming non municipalities were deemed but one jurisdiction in King, we have a highway cut between Renton and Newcastle, Renton and the unincorporated area,  Maple Valley and the unincorporated area, and then road cuts however counted where the unincorporated area is chopped, which is any pavement correct? That would be a nightmare to count potentially. And you think tie breaking preference rules are complex and hard to apply (which even someone as obtuse as I think are quite simple and straight forward)!

Anyway, the incentive would be to have the CD line follow the Cedar River to minimize such cuts I suppose, and adding a road cut for Black Diamond, per the map below. Although you might say Black Diamond is not connected by a state highway so that is off the table, but I don't agree with that rule either with respect to internal cuts in a county. Any pavement will do. Anyway, it seems the the road cuts internally are quite low.

Of course, with my bridge chop definition and rule, the posited Pub temptation that is alleged to be doable with a winning score would be off the table, but I digress. It appears that we may permanently part ways on the bridge chop issue. Pity.

Actually I thought this road cut count was as simple as could be. A regional connection exists between counties where there is a numbered state highway. Within a county there is a single connection between each locally connected fragment. In a macrochopped county there is a connection between counties and fragments where a regional connection exists (used for the second number). One then proceeds to count the cuts as I did above (number after the slash).

Is there an instance in my description above where this didn't work?

Regional connections within counties are numbered county highways? In King, some roads without numbers seem like arterials, and given more prominent display on the DRA than others, which look like residential streets. I don't see any numbered county highways actually. Anyway, it seems like my map minimizes such cuts. And within the unincorporated area, assuming no subunits, I assume one counts each place where a highway however defined is cut.

[muon2]

I understand all of the above, but I don't think it is responsive to my post. Of course, we are only talking only about macro chopped counties. I accept that the precinct nesting issue should not be the tail that wags the dog. It is just an indication of whether or not there are subunits that anybody cares about for election purposes. The issue is where there are no subunits, how necessary is it really to artificially generate them?  You say one reason is that you might end up with wide partisan swings. I reply that maybe that is not so bad, if it gets the SKEW down, provided the maps don't get too erose looking. You offered up a map that looked erose, but much of that is really due to the shape around the Newcastle area, with a highway squeezing by to keep the CD connected. I grant it looks kind of ugly however. I put up a map that was not so ugly, and also did not impact SKEW, that would seem to me to have fewer highway cuts. You assert that there are more. I don't understand how. I am not sure how road cuts are counted between subunits in macro chopped counties. I thought it was any road. Granted if there is one big swath of land in King without subunits it gets to be cumbersome to count every road, where the unincorporated subunit is chopped. The Cedar River however, happens to make it all manageable in this case it would seem, particularly if any pavement connections will do within a county.

Speaking about chops of subunits in macro chopped counties, I am thinking of allowing one chop without penalty of a subunit to get the SKEW down, subject of course to a road cut count, so it does not get too ugly. The same thing within a big city. It if has neighborhoods, allow one chop to get the SKEW down, subject to the erosity rules. You won't like that one either probably. And what do you do with cities that have no neighborhoods, like seemingly Tucson, that are usable? What about Phoenix? Just create some out of whole cloth? What I do is try to have straight lines preferably following highways. Generally, in cities without apparent usable neighborhoods my chops within cities are highly "artistic" and pleasing to the eye. I like it that way. Anyway, the point being, that if there are going to be subunits everywhere in counties subject to macrochops, generated somehow, the one bite hope rule might be a workaround to deal with SKEW issues. That might be the right compromise, the Golden Mean, to resolve this particular food fight. Dogs get one bite with a negligence rap, and maps similarly get one bite in this context.

The big question you raise, and it was one that we settled two years ago over MI, is should political measures be used as a filter for maps or for informational purposes only. Two years ago your feeling matched the consensus that it should not, so I provided those scores on submissions from train, jimrtex, you and I, but did not cut on them. We used them to see if other rules were gaming the system. Now you seem to revised your position, but it will need serious convincing for me that will include a complete revisit of our work in MI.

I'd like to put off the SKEW question here and focus on how to measure erosity in macrochopped counties with no agreed subunits. The erosity metric assumes established connections that can be broken, and that in turn assumes that every parcel is assigned to something that can be connected. We have no agreement on assignment for unincorporated parcels, so they have no connections, so there is no erosity.

As I look at this I would create a hybrid of munis and school districts extending out from their munis where munis are embedded. That would comport with the public sense of belonging that jimrtex favors and I think helps sell the concept, but maintains smaller munis as both you and I would use as natural divisions. My experience with county election boards is that they don't care much about the districts not under their purview unless compelled by state law. Hence munis are treated well, but school districts are not.

For the large cities I would do what I did with Detroit and the VA cities. I would look at each city web site to find their planning areas. They almost always have them. Those then become the subareas within the city.

[Torie]

You could create an erosity test by just counting the highway cuts, however defined, within the cut "jurisdiction," in this case being the unincorporated area. That seems quite logical to me. You don't count highway cuts at all in chopped jurisdictions? If you don't, then you can draw the lines anyway you want in the chopped jurisdiction it would seem. What if a school district takes in a city and territory outside it? Is the territory outside the city its own jurisdiction for your purposes?  

I am really not that hostile to creating such artificial jurisdictions, if something can be created that is workable. My fix that I think will work is my one chop rule. It is not only about erosity, and SKEW per se, which I think are very important, but also that map lines in this context will be a function of population accident if there are lines that happen to avoid any chops. That is a horrible result in this context - just horrible.

You raise the estoppel argument often with me. I don't feel estopped at all. If I have new insights, as I gain more experience and knowledge, I am going to change my mind, or finally form an opinion, or an informed opinion. In some cases, I really didn't understand some of the details, and certainly not all of the implications. This for me, will always be a work in progress, always, until I am satisfied that the best approach possible can be fashioned.

Hopefully at the end, where we cannot agree, it will be after thorough discussion, with the issues properly hashed out, and the competing considerations and plusses and minuses, understood. Then at least where we disagree, it is because we weight the competing considerations differently. If you have informed commentary down the line, you are going to face these questions again, and again, and again. You might as well face them now.

And as I said, your big opposition is going to be from Democrats, once they understand the game. Your system, if it remains rigid, will tend to lock in a Pub majority at present. The Dems are better off with commissions that they can game, with pseudo non partisan human tie breakers like Mathis, or the dumb Pubs on the CA commission, some of whom were probably false flag or close to it, and BS COI chat, not something that a computer does, that will screw them on a nationwide basis, after they pick up their ten or so seats net. At least my one bite rule, provides more flexibility to give them hope. Drawing that second Dem CD in Phoenix is really almost mandatory, assuming the lines don't look too erose. It really is. I don't want a neighborhood cut that could be avoided to interfere with that. It's one thing to have which counties go together be a population accident. It's another for which neighborhoods and so forth go together be a population accident. Your reply? Well the highest scoring map need not be picked! I understand. But lower scoring maps have a headwind against them. And they might be knocked out of the box entirely if not pareto optimal.

I might add that my bridge chop rule I think will tend to hurt the Pubs. They can't without penalty I in this case losing by virtue of the preference rule) suck up rural fragments in two counties in some cases by applying it, as was true here in Washington. Sucking up two more rural fragments is almost the definition of a Pub gerrymander. Instead it is the other CD in the urban county that will suck up the rural fragment in the adjacent county. The two rural fragments would tend to be put into different CD's.

The planning areas in Tucson are not really usable, given how the roads flow. Take a look.

[muon2]

You could create an erosity test by just counting the highway cuts, however defined, within the cut "jurisdiction," in this case being the unincorporated area. That seems quite logical to me. You don't count highway cuts at all in chopped jurisdictions? If you don't, then you can draw the lines anyway you want in the chopped jurisdiction it would seem. What if a school district takes in a city and territory outside it? Is the territory outside the city its own jurisdiction for your purposes?  

I am really not that hostile to creating such artificial jurisdictions, if something can be created that is workable. My fix that I think will work is my one chop rule. It is not only about erosity, and SKEW per se, which I think are very important, but also that map lines in this context will be a function of population accident if there are lines that happen to avoid any chops. That is a horrible result in this context - just horrible.

You raise the estoppel argument often with me. I don't feel estopped at all. If I have new insights, as I gain more experience and knowledge, I am going to change my mind, or finally form an opinion, or an informed opinion. In some cases, I really didn't understand some of the details, and certainly not all of the implications. This for me, will always be a work in progress, always, until I am satisfied that the best approach possible can be fashioned.

It's that "however defined" that is the crux here. My only constraint is that a macrochopped county have subunits that include every parcel. I assume that they will not be the same in every state, since potential county subdivisions vary so widely across the 50 states. For 25 of them we have the town/township model. Each of the others needs its own definition.

In WA I think precincts are too small and using whatever chop pieces emerge are too big (that's the reason for my King gerrymander - I think it should be disfavored on the EROSITY score). Let's stipulate that the munis are separate from each other in subunits. The question is how to group the remaining unincorporated parcels. This can be done by clustering them as psuedo munis, perhaps by school district. In some cases the pieces may be so small that it makes sense to aggregate them with the associated muni, but in other places they would not be.

I think the reason for my estoppel arguments comes from my science background. If I build a model that fits one set of data, I'm very interested in testing the model with new data. Important refinements to the model may result. However, if there is a significant change to the model I feel compelled to double back to my original data set and see what impact it would have. I cannot rely on it working on the new data alone. If the choice is between an old model that works for one set of data and a new model that works for a different set of data, my training is to stay with the established model until the new model is shown to fit the old data, too.

[Torie]

Obviously that is what one is left with, if you want subunits. Find something, and if it does not work, then modify what you find selectively, based on stuff being too small or whatever. The subunits also need to work with state highways, ala Tucson, or something will collapse. Perhaps the one chop rule will help there. And what if the subunits are too erose? Keep looking for others that are less so, until one finds something?

I really don't think my proposed rule changes are going to cause a collapse of your maps. They are changes on the margins. With total respect, I consider that argument make weight. I have yet to see one example of that anywhere. Well there is your King County map perhaps, if it is the winning map, which I am not persuaded it is, but I agree that if there were no subdivisions, you need a road cut regime, and I further agree that it is not good to have some counties with no subunits ala King, and other large counties that do have subunits, if it can reasonably be avoided. I am not against subunits per se. I think I can help SKEW with my one chop rule, at least at the margins (with that chop leashed by road cuts however defined within the chop, and bearing in mind, that with relatively small subunits, we are not talking about chops of areas with big populations. It just avoids situations where there is a map that happens to have no chops, that then knocks all the other maps out).

The other argument you make, is that my changes are too complex. That one brings a smile to my face, given many of the rules that have been fashioned. I really don't think preference rules, or the one chop rule to help with SKEW, are particularly complex at all. I mean, how can it be, when I actually understand it!  The nick rule does add some complexity I must admit. But if it causes better maps (by lower chop and UCC penalty points), it is probably worth it. It won't really apply very often anyway.

I still don't know what the regional roads are in King County by the way, that has no county road numbers it seems, assuming that is what regional roads mean. I guess that point is moot, except perhaps within chopped subunits, if one has subunits. Well not really I guess, assuming that there is no numbered county highway between subunits intra-county (if that is the definition of an intra county regional highway), unless then any pavement counts as a highway cut. And we don't want there to be too many such highways, or the map is back in the box again, with only one map to be drawn even with the one chop rule. We want some sensitive highways, but not too many. In short, we want a regime that helps with SKEW, but still has maps that look reasonable to the eye. It's another Goldilocks situation.

By the way, I assume that if you chop into a city, and then also a neighborhood within the city, that  counts as two chops, right?

Macro-chops are hell. A necessary hell, but still hell. It is by far the toughest aspect of this all - at least to me.

[jimrtex]

But I will get into that dealing with your King County map on that thread. On that map, did you count non state highway road cuts between municipalities in separate CD's that do not involve another county? I ask, because in general, the county chops are incentivized to go where fewer jurisdictions exist, which would tend to help the Dems because it is less populated areas that are chopped. In general, if one CD takes in the more rural areas, and the other CD takes in the populated areas, you will have one CD appending inside a county a ton of localities, with a lot of road cuts. Each and every road that leads out from a municipality to the balance of a county, even if there are no sub jurisdictions in the balance of the county, would count as a road cut. So that issue needs to be resolved first.

I'll reply over here.

The method I used to assess erosity in the King example was the one that was initially developed and applied in our exercise in AL and other states back in 2013. That only counted at most one connection between pieces of different districts within a county. We thought about applying townships in all cases, but an initial look at minor chops in OH showed that counting all the connections between townships/munis in a chopped county overly penalized those minor chops. However, something was needed to deal with dense counties like Cuyahoga where erosity could only be measured in terms of adjacent munis.

When we came to MI we dove into detail on when those connections between munis should apply. The preferred metric was to define a macrochop then apply the subunit connections only in those counties subject to a macrochop.

The essence of this erosity measure relies on having every parcel assigned to a division. For example every parcel in a state is in a county or equivalent so erosity at the county level makes sense. In MI every parcel in a county is in a township or muni so again the metric works appropriately. What we are trying to do is accommodate macrochopped counties in states where not every parcel is in a township/muni.

A big complication is that you only want subunits that reflect preexisting precinct lines. However, as I look at various state statutes I see it's usually the opposite from the direction that would work best for you. The precincts are updated to conform with political units after redistricting, not before. The policy reason is to reflect any new chops with new precinct lines.

In King there are munis but not every parcel is in one. I did not consider muni-muni erosity because I don't have a metric to define a connection when there is an unincorporated gap. For example there is an unincorporated gap between Kirkland and Bothell on my map. The gap was small enough that today they are contiguous, though they weren't in 2010. Either way, a meaningful erosity needs to assign that 2010 unincorporated area to a subunit.

Who wants to use precinct lines?  Torie, or a representative mapdrawer?

Why wouldn't it be better to expect that King County would define county subareas well in advance of the redistricting process?

[muon2]

Car rides are good for thinking, but less so for drawing, so this took a while. Here is a rule based way to create subunits, applied to King. These are the rules I used:

1. Each incorporated municipality is in its own subunit.

2. Contiguous unincorporated areas in the same school district are in a single subunit. Precincts that overlap multiple districts are assigned to the district with the most voters in that precinct.

3. Incorporated municipalities entirely within an unincorporated subunit are merged.

4. Unincorporated subunits entirely within an incorporated municipality are merged.

5. Unincorporated subunits smaller than 0.5% of the quota are merged with an adjoining municipality in the same school district.

I ended up with 51 subunits shown below as a whole county and with detail on the western side. Colors are grouped by school district with darker shade for munis and lighter shades for unincorporated subunits. If this seems workable I can produce the connection map to use for erosity. I also have the populations of each of the subunits if they are of interest.

[Torie]

What is the purpose of rule number 5? Is this a case where the portion of an unincorporated area that is not within a municipality but is within the same school district is less than 0.5% of the quota?

[muon2]

What is the purpose of rule number 5? Is this a case where the portion of an unincorporated area that is not within a municipality but is within the same school district is less than 0.5% of the quota?

Yes. There are disconnected fragments outside of munis that often are assembled into a precinct. They are small and pose connection problems. Given the small pop (microchop size) it is much easier to append them to something larger.  The munis already pick up pop from surrounded fragments so picking up these other isolated fragments is consistent.

[Torie]

OK. You might clarify in your text that that is your intent.

I have been busy too.  Here is the neighborhood map for Phoenix.



Of course the precincts are not nested. Here is my best effort of approximating the neighborhoods.   Note the wide range in populations. Maybe the tiny hoods on the northside should be combined, maybe not. I am not sure it will make much difference.  Tiny hoods unless strewn around randomly, will tend to be in one CD anyway.


 
As feared, the other version map version chopping Phoenix, which shoves AZ-09 three points in the Pub direction, and into Pub hands, seems to have the same number of cuts.  I am counting for intra county cuts, any pavement. I don’t think picking and choosing what is a qualifying road is wise. Just like when a particular division of a city or county avoids a chop, based on the accident of what population fits, I think that is arbitrary, and interferes with SKEW patrol. As long as the map does not get too erose, that is bad policy. It will never be accepted I don’t think, and it should not be accepted. It is one thing to have such arbitrary and accidental results with county chops, because given our UCC rules, just how the rural counties are shoved around, will rarely make much partisan difference. That is not true when handling macro chopped counties. I might add that the pro Pub map version of Phoenix gets close to avoiding a neighborhood chop. It is off by maybe 15,000 people. If the north end of Phoenix grows faster than the south end, that discrepancy might disappear. It is a roll of the population dice, in a context where the dice should not be rolled, assuming erosity is kept under control.

 




So I propose for marcro-chop situations the one bite rule. You can have one extra intra county chop over the minimum that can otherwise be achieved while hewing to the VRA, without penalty. The map with the fewest road cuts, counting any pavement as an eligible highway, would still prevail however if the chop count is the same after getting your one bite. In this case, the two maps are tied on road cuts in Phoenix, so SKEW is the tie breaker, and the more Dem friendly map prevails – as it should. And it should prevail even if the Pub map managed to avoid that chop of Deer Valley on the far north end thereof.

I might add that we have never really thoroughly explored this issue before. Most cites, and for that matter, urban counties, of the size that brings this issue into play, with which we have dealt, are either so pro Dem, or driven by the VRA, or both, that it makes no difference. But it makes a difference in Arizona, and Washington potentially, and maybe in the Dallas and Houston area, to name some areas off the top of my head. Thus even if estoppel were an appropriate response in some contexts (I don't think that appropriate in any context as to the metrics here), it certainly is not in this context.

So here is the list that I can think of, which are sources of disagreement (maybe reconcilable or maybe not) or under discussion, or works in progress.

1.   Definition of bridge chops, and the penalty regime therefor.

2.   The highway nick issue.

3.   The issue of whether you need to chop into counties on state highways and the penalty therefor.

4.   What counts as an intra county highway cut.

5.   The one bite rule.

6.     The regime for handling Indian Reservations. My statutory text is substantively different from Muon2's. I oppose treating reservations in a way that is identical way to counties. They are hybrids, and should be deemed so. I also oppose any incentive to have two adjacent reservations of different tribes, being placed in one CD. The result is to reduce flexibility, and incentivize chopping counties to bits. I think my proposal is the right balance.

I have given up on the precinct nesting thing, and I believe that the one bite rule, makes Muon2’s protocol of shoving everything into districts within a county a tolerable constraint, vis a vis the SKEW patrol issue.

Oh, on the matter of where municipal lines are (the Maricopa area on its west end is a mess, I found this utility. You’re welcome! ☺  

Notice, that my perusing of the muni lines in west Maricopa using this utility, caused my map to go ugly in that area, and that Goodyear is chopped. That is another problem with the macro-chop rules. It tends to incentivize incurring pack penalties, and maybe both pack and cover penalties. The dice are loaded against macro-chops.  Which is yet another reason to go with the one bite rule to mitigate that.  It’s one thing to stack the dice, it’s another thing to load them. The one bite rule helps to get from the load category to the stack category.

[muon2]

I'm not sure I'm clear on the one bite rule. Is this for cases that are macrochops and governs the number of chopped subunits? If so, is this applicable in well-divided counties like in the Midwest? How does it function if there are 3 or more districts in a macrochopped county?

I think you misunderstood my reservation rule. Each separate reservation was its own entity, much like my split of King unincorporated areas by school district.

Keep in mind (and it may already be so) that inequality, chop and erosity are used to form the Pareto set. Skew (and polarization) comes in afterward, applied to those plans in the Pareto set.

[Torie]

I'm not sure I'm clear on the one bite rule. Is this for cases that are macrochops and governs the number of chopped subunits? If so, is this applicable in well-divided counties like in the Midwest? How does it function if there are 3 or more districts in a macrochopped county?

Yes, you get one extra chop of a subunit in macro chopped counties, and it is per each set of two adjacent CD's. There is no one bite rule otherwise. It is not needed, and population accidents will be accepted, because most are into limiting county chops. By the way, we need a rule as to when a subunit gets too large, so one needs to create neighborhood subunits. Perhaps it should obtain where its population size is in excess of two microchops, which would be around 70,000 people. Maybe that is too small. Some of the hoods in Phoenix are more like 150,000-200,000 people.

Ideally, rules should be the same across the Fruited Plain, absent a compelling reason to depart from that. In the Midwest, it will not matter, given that we have done maps for all the states there, and know the lay of the partisan, racial, and population land. But the same issue obtains, if it did matter.

I think you misunderstood my reservation rule. Each separate reservation was its own entity, much like my split of King unincorporated areas by school district.

OK, that resolves one area of discrepancy between our two rules. Your text did not read that way, the way I interpreted it. Your regime still requires county chops however, rather than just a preference. I don't like that, and I don't think it is justifiable. Bear in mind that most reservations are small in population. Avoiding a chop of such tiny entities, causing a map potentially to change substantially, is a really egregious example of the tail wagging the dog. If say, by mandating a chop of Pinal County, the CD population were too small for the Pinal County CD, then potentially the whole map goes down the drain, or if a Maricopa CD needed to chop out. The preference rule avoids such unpleasantness.

Oh, I remember the German guy's name now. We are talking about the Lewis Trondheim rule.

Keep in mind (and it may already be so) that inequality, chop and erosity are used to form the Pareto set. Skew (and polarization) comes in afterward, applied to those plans in the Pareto set.

I always bear that in mind, each and every second of my time while awake, and sometimes even in my dreams. But here's the thing. My AZ maps are identical, except in Phoenix. So guess what? One of the maps, unless otherwise tied (which fortunately appears to be the case here) absent my one bite rule, is not on the pareto optimal frontier. It needs to be the map that has the lowest SKEW, and not driven by population count accidents of the respective city hoods. The one bite rule will not cause ugly, erose maps.

You have inequality above SKEW? Not good. It should be below SKEW, not above. Inequality is a trivial concern in the larger scheme of things. Add that one to the disagreement list!

The tie breaking order should be:

1. Bridge chops
2. Indian reservations
3. SKEW
4. POLARIZATION
5. Using state highways for chops into counties as opposed to any pavement, to wit, state highways, then county numbered roads, and then any all weather two lane highway pavement
6. Inequality

[muon2]

Let's set aside our disagreements about chop penalties and the overall scoring metric for the time being and return to the purpose of this thread which is defining the EROSITY score. It is one of the axes of the Pareto diagram, and isn't directly impacted by how one scores chops, skew, or the rest.

Here is my working definition that I posted earlier in the thread.

A political unit can be represented by a node that is the political center of that unit. For a county the node is the county office where the elected officials meet. For a city or town the node is the city or town hall. For a precinct the node is the polling place. Units are connected based on the path that connects their nodes.

Two units are locally connected if there is a continuous path of public roads that allow one to travel between the two nodes without entering any other unit. Local connections can include seasonal public roads. A local connection path can be traced over water without a bridge if there is a publicly available ferry that provides part of the connection. Units smaller than a county must be locally connected within a district.

Two counties are regionally connected if there is a continuous path of numbered state or federal highways that allow one to travel between two nodes without entering any other county. If a node is not on a numbered highway, then the connection is measured from the point of the nearest numbered highway to the node. The path may only use roads that are generally available all year. Regularly scheduled year-round ferry service may be included in the path of a regional connection. Counties must be regionally connected to be connected, except that counties within a cluster are connected if they are locally connected.

There is often more than one possible path to connect to nodes. For both local and regional connections the connection between two units is considered to be based on the path that takes the shortest time as determined by generally available mapping software.

It is on the definition of regionally connected that there is still disagreement on the corner cut issue. Is the rest is agreed?

Now let me put up my working definitions of the impact of chops on connections.

A chop occurs when a geographic unit is divided between two districts. For congressional districts, chops less than 5% of the quota are simple chops. Those chops in excess of 5% are macrochops.

For simple chops each connection from the original unit to adjacent nodes is assigned to one of the chopped parts. The part that has the border containing the connection path is assigned the connection. Note that connections to adjacent nodes cannot be lost by chopping a unit. Within the chopped unit the chopped parts are connected to each other if they are locally connected.


The best example is the one I used for King earlier in this thread. The first number before the slash for each district followed the rule above. A modification to the above would require that each chopped part have regional connections to adjacent county nodes that could be cut. That is what went into the second number after the slash in the King example.

The rule is consistent with our methodology in the MI exercise. Note that the only increase in erosity compared to keeping the county whole occurs from the severed internal link, and can reduce erosity as we saw with the recent NC plans.

The modified rule tends to increase erosity for simple chops, and the thinking two years ago was that the chop penalty was sufficient, without adding additional penalties for erosity.

For macrochops each unit is replaced by its complete set of subunits. Subunits are connected within a county if they are locally connected. Subunits are connected across a county line if they are regionally connected.

EROSITY is the sum of all connections severed by the boundaries between districts.

I think that the concern about the type of connections in Phoenix is solved by using the local connection requirement. The type of road doesn't matter in that case.

The principle issue here I believe is what constitutes the subunits of a county in a state with no township level government, and what constitutes subunits of a large city that will typically be chopped. Both the King county and Phoenix discussions go to that issue. We had similar discussion for Detroit in the MI exercise.