Using Urban County Clusters To Guide Redistricting

[Torie]

As to your maps Train, all are legal. On its face, that erose black CD in Mississippi might be questionable, but it is legal because a less erose black CD can be drawn, and the only reason it was not, was not a racial motivation, but an urban cluster one.

I think the VRA would clearly require the non erose black CD in Mississippi, assuming the black population is contiguous. You just took one whole urban county and put it with some rural counties. The VRA is unfamiliar with multi county urban clusters.  

The Alabama CD is the tough one, as to whether the VRA requires it. The odds are higher if the black population is reasonably contiguous, stretching from the black Birmingham chop (not a whole county) down the state. If not contiguous, and the CD just picks up black Birmingham via the chop, and then goes through a sea of whites to pick up some disparate black rural counties at the southern end, the odds are much less that is required, and in fact, it probably isn't.  Flip a coin as to whether one needs to do one urban county chop, and append it to a bunch of contiguous minority populated rural counties where one does not have to wade through a sea of whites. There it might depend on erosity. If a non erose CD can be drawn, like you did, then you have to draw it, if not, you don't. I don't think the VRA demands erose CD's under any circumstances (except perhaps within an urban area).

[traininthedistance]

As to your maps Train, all are legal. On its face, that erose black CD in Mississippi might be questionable, but it is legal because a less erose black CD can be drawn, and the only reason it was not, was not a racial motivation, but an urban cluster one.

I think the VRA would clearly require the non erose black CD in Mississippi, assuming the black population is contiguous. You just took one whole urban county and put it with some rural counties. The VRA is unfamiliar with multi county urban clusters.  

The Alabama CD is the tough one, as to whether the VRA requires it. The odds are higher if the black population is reasonably contiguous, stretching from the black Birmingham chop (not a whole county) down the state. If not contiguous, and the CD just picks up black Birmingham via the chop, and then goes through a sea of whites to pick up some disparate black rural counties at the southern end, the odds are much less that is required, and in fact, it probably isn't.  Flip a coin as to whether one needs to do one urban county chop, and append it to a bunch of contiguous minority populated rural counties where one does not have to wade through a sea of whites. There it might depend on erosity. If a non erose CD can be drawn, like you did, then you have to draw it, if not, you don't. I don't think the VRA demands erose CD's under any circumstances (except perhaps within an urban area).

I'd say that the black district in my Mississippi map is not erose in the least; what is arguably erose (and even then, it's not really erose by the standards of maps that get drawn) is MS-3, the district in between the black district and the southern MS district; and the existence of a VRA district in Jackson and the Delta, plus a population distribution that puts most of a district in the panhandle + Hattiesburg, requires it.  If you still consider my MS-2 to be erose, I would like your technical definition of erosity, for I'm not seeing it.

I think it's pretty odd that you'd expect an erose CD to be required within an urban area, but not required outside of one.  What's the justification for that?

Also, you keep saying "sea of whites" and I'm wondering what exactly is the threshold for a "sea"?  The Alabama CD does in fact go through a couple white precincts, in the Tuscaloosa area and on the edge of Jefferson County, but the total population in those precincts is a vanishingly small percentage of the total population of the district- if one were to excise northern Tuscaloosa, which is only there for county integrity purposes, it would be probably about 20K people at most.  I feel you're using a pretty loosey-goosey term here, and would like to know if there is a more exacting standard you would strive for.

[Torie]

Contiguity is a function of population and distance. If those 20K whites took up the middle half of the CD, there is no contiguity. The way you described it, there probably is. The way to know, or know better, is to look at the DRA racial map, and see if it generally has shades of blue in that area. I consider TX-23 as having a contiguous Hispanic population.

It is just my instinct about intra urban cluster erosity, and whole county erosity. In whole county erosity, the distances are longer, and there is less of a community of interest. But in a city, maybe the Hispanics live along various railroad tracks, strung out all over, or on the flood plain that runs up a river, so when you draw their CD, yes it is erose. But is it a community of interest? You betcha. It's just that the cheaper housing locations are erose, but they all go to the same places day in and day out, the same churches, and so forth.

I was talking about the green CD below. I consider it erose, and would never personally draw it unless I had to. Your other black CD on the other MISS map was not erose.

[traininthedistance]

Contiguity is a function of population and distance. If those 20K whites took up the middle half of the CD, there is no contiguity. The way you described it, there probably is. The way to know, or know better, is to look at the DRA racial map, and see if it generally has shades of blue in that area. I consider TX-23 as having a contiguous Hispanic population.

It is just my instinct about intra urban cluster erosity, and whole county erosity. In whole county erosity, the distances are longer, and there is less of a community of interest. But in a city, maybe the Hispanics live along various railroad tracks, strung out all over, or on the flood plain that runs up a river, so when you draw their CD, yes it is erose. But is it a community of interest? You betcha. It's just that the cheaper housing locations are erose, but they all go to the same places day in and day out, the same churches, and so forth.

I was talking about the green CD below. I consider it erose, and would never personally draw it unless I had to. Your other black CD on the other MISS map was not erose.

Oh, yeah, that one is erose as all f[inks] and I just put it up there as a spitballing proof-of-concept.  The other MS-2 would definitely be preferred.

As for the Alabama map, this is that AL-7 minus 55K whites in the northern half of Tuscaloosa County (again, just there for county integrity, not necessary otherwise):



And then we take away the actual bridge, which turns out to be more like 36K residents rather than 20K, mea culpa:



No, it is not *strictly* contiguous.  But *strict* contiguity is not to be found even in TX-23, honestly (a lot of that area is Anglo-majority "mixed" precincts, and it is all so empty besides).  And it is not like the entirety of Tuscaloosa County is lilywhite; the remainder of Tuscaloosa County on the last map is over half the county's population, and is majority-white but not by much at all (55% White/40% Black).  Tuscaloosa County is, as a whole, 68W/27B.  

While I suspect your opinion may differ, I think it ought to count, and would expect most judges to agree.

[traininthedistance]

It also has just occurred to me now, because I'm an idiot, that:

The concept of Urban County Clusters is to handle big blobs of population which are difficult if not impossible to divide into equal population districts based on county boundaries, or which may result in perverse outcomes when they do - such as a large suburban county with the population for half a district stretching across the state to pick up the rest of the district.  In addition in urban areas, county boundaries are a weaker source of identification, with residents, workers, and shoppers crossing the county boundary oblivious to its existence.

An Urban County Cluster means "county-based districts don't work as well here, identify a compact area that contains the blob and can be represented by one or more districts, and treat them separately.

is actually saying that county chops should not really be minded so much within the cluster (perhaps intra-cluster chops should be counted at a discount rate, if at all, in any scoring system?), and that perhaps within the state as a whole the UCC is to be treated as a super-county for the purposes of chop identification and minimization.

Which is a stance I wholeheartedly endorse.

Within major metropolitan areas, you will probably have to cut counties, and the magnitude of the county fragments may be large.  I can't really say that splitting County A 300 thousand and 200 thousand is worse than cutting County B 300 thousand and 125 thousand, along with associated grouping of other counties is better or worse.

I see the division of multi-county regions as a separate step, and it can be controlled by limiting the number of cuts to the minimum necessary (N-1), disallowing multi-spanning where two or more districts have parts of the same pair of counties, and emphasizing respect for local communities of interest, particularly where there are local political boundaries available.

Note that I specifically permit forming whole county regions within UCC.  That actually eliminates a (significant) county cut.

In New Orleans, Jefferson+Orleans (1.026), Jefferson+St. Tammany+St Charles+St John the Baptist (1.013), Jefferson+St.Tammany+St.Charles+Plaquemines (0.980), and Jefferson+St.Tammany+Plaquemines+St.Bernard (0.958) are all within a 5% limit.  However, they both  cut off the remnant of the UCC cluster from the rest of the state.

There are cases, to be clear, where local sentiment/communities of interest will in fact prefer multi-spanning of districts.  The canonical example is on Long Island, where the new (and widely praised) court map has two Nassau-Suffolk districts, one each for the North Shore and the South Shore.  IIRC the geography of Washington State seems to suggest that double spanning would be appropriate there as well, perhaps to allow an eastern King district to spill over to Snohomish and/or Pierce, as well as one in the denser western edge. 

I can understand why one would want a rule against it anyway, but we should recognize that it is not completely without its drawbacks.

And of course I absolutely agree that trying to hew to local political boundaries should be a priority in urban areas.

[muon2]

Here are the rules I thought we were using for UCCs. The trims represent the changes to jimrtex's initial set of maps on the thread.

These changes are based on the following definition:

County is (1) constituent of a metropolitan statistical area; (2) an urbanized area is the most populous urban area in the county; (3) and 50% or more of the county population is in urban areas (alternatively, less than half the county's population is rural.

Thus qualifying counties are either (a) a central county of the MSA (1 and 2 are necessary conditions) and have a mostly urban population; or (b) an outlying county, that had been initially identified as a central county of a metropolitan statistical area, but that had been merged into another CBSA based on commuting, and has a mostly urban population.

Counties of type (b) had either initially been identified by a pseudo core (eg South Lyon-Howell UA) or a weak core (eg Port Huron UA). 

Counties that are excluded by the 50% urban requirement were misidentified as urbanized counties based on the Census Bureau using a single criteria for both Metropolitan and Micropolitan statistical areas (ie 5,000 persons in the county in the largest urbanized area).   5,000 persons is between 10% and 50% of an urban cluster that is a core.  It is less than 10% of an urbanized area that is a core, and thus some counties that are fairly peripheral to the urbanized area, are classified as being "central" to the metropolitan area.  These counties are really bedroom counties, but have some people sleeping on the couch in the living room.

Alabama:
Montgomery (trim Elmore) 46% urban.

Arkansas:
Fort Smith (trim Crawford) 48% urban.

Connecticut:
Torrington (eliminate Litchfield).  Bridgeport UA, Danbury UA, Hartford UA, and Waterbury UA come into Litchfield County, but Torrington UC is the most populous urban area.  In part because of multiple target counties for commuting, Litchfield is not an outlying county of any.

Delaware:
Salisbury, MD (add Sussex)  Salisbury, MD urbanized area is the largest urban area, in a county with a bunch of small urban clusters, such that Sussex County is 59% urban.

Georgia:
Athens (trim Oconee) 49.6% urban.
Dalton (trim Murray)  30% urban.   I had misstated that Dalton was in Chattanooga, TN MSA.

Indiana:
Terre Haute (trim Clay) 39% urban.

Kansas:
St.Joseph, MO (trim Doniphan) 30% urban.  The portion of St. Joseph UA in Doniphan is so small (2368) it does not even qualify as a central county under the Census Bureau definition.

Kentucky:
Clarksville, TN (trim Christian) Hopkinsville UC is largest UA in county.

Louisiana:
Baton Rouge (trim Iberville) 41% urban.

Maine:
Portland (trim York) 43% urban. 
Bangor (eliminate Penobscot) 42% urban.

Maryland:
Balltimore (trim Queen Anne's) 46% urban.
California-Lexington Park (eliminate St.Mary's) 49.6% urban.

Michigan:
Lansing (trim Clinton) 47% urban.

Minnesota:
La Crosse, WI (trim Houston) 43% urban.

Mississippi:
Hattiesburg (trim Lamar) 49.6% urban.

North Carolina:
Asheville (trim Haywood) 45% urban.

Ohio:
Wheeling, WV (trim Belmont) 45% urban.

Oklahoma:
Oklahoma City (trim Logan) 45% urban.
Tulsa (trim Creek)  46% urban.

Pennsylvania:
Bloomsburg (trim Montour) 46% urban.
Gettysburg (eliminate Adams) 46% urban Hanover Urbanized Area qualifies the county as a central county of a metropolitan statistical area, but Hanover is in York County with a long flanking arm down the York Pike into Gettysburg.   York (City) Urbanized Area is the much larger urban area that qualifies York County.  York-Hanover MSA consists of York County, while Gettysburg MSA consists of Adams County (MSA are named based on the largest city, regardless of the urban area that forms the core).

South Dakota:
Sioux City, IA (trim Union) 39% urban.

Tennessee:
Kingsport (trim Hawkins) 42% urban.

Texas:
Beaumont (trim Hardin) 48% urban.

Virginia:
I had to restore 3 independent cities to their original county to maintain the counties as part of the urban county cluster.  I suspect that best solution would be to treat all IC as part of their original counties.

Kingsport-Bristol-Bristol, TN-VA (trim Washington + Bristol) 46% urban.
Washington, DC (trim Fauqier) 43% urban.
Lynchburg (trim Amherst) 36% urban.
Lynchburg IC added to Campbell to maintain its status.
Petersburg IC added to Dinwiddie to maintain its status as part of Richmond cluster.
Hopewell IC added to Prince George to maintain its status as part of Richmond cluster.
Roanoke (trim Botetourt County) 36% urban.

West Virginia:
Huntington (trim Wayne) 35% urban.
Beckley (trim Fayette) 42% urban.

[Torie]

I believe it quite unlikely Train's black Alabama CD is required by the VRA. To get to Birmingham is a bridge too far. If it is not a bridge too far, than a fortiori, Muon2's NO to BR affair is also required by the VRA. I don't think so.



Perhaps a closer case is the black Bama CD below (it's 49.9& BVAP, but with a change of a precinct's boundaries in Tuscaloosa, can be pushed up to 50%). It has a reasonably contiguous black population, is not hideously erose, but does wander, requires strategic chops, including of a municipality in Autuaga County and the stab grab of a chopped Tuscaloosa municipality), and makes a hash out of the two CD's to the south. (It also chops the Montgomery urban cluster, but I don't think the VRA cares too much per se about urban clusters, although it might take cognizance of them if the state does up to a point.) That CD would make an interesting case, but if I had to bet, it is more likely than not, not required either.

[Torie]

So Jimtex, to cut to the chase, the census bureau identifies the urban percentages in counties, and you excise those that are included in a census urban cluster, but have less than 50% of the population as urban. Is that correct?

[traininthedistance]

I believe it quite unlikely Train's black Alabama CD is required by the VRA. To get to Birmingham is a bridge too far. If it is not a bridge too far, than a fortiori, Muon2's NO to BR affair is also required by the VRA. I don't think so.



Perhaps a closer case is the black Bama CD below (it's 49.9& BVAP, but with a change of a precinct's boundaries in Tuscaloosa, can be pushed up to 50%). It has a reasonably contiguous black population, is not hideously erose, but does wander, require strategic and excess chops (including the stab grab of Tuscaloosa), and makes a hash out of the two CD's to the south. That CD would make an interesting case, but if I had to bet, it is more likely than not, not required either.

I think what this exchange is proving, more than anything, is that your idealistic intuition is being dashed among the reality rocks of rural minority populations, and is just not handling them well at all.  I don't think there's any way in hell that something like the lower map wouldn't be required (I played around with similar things and was able to get up to about 50.5% BVAP without appreciably more erosity).  The Mobile CD is not made a "hash of", either. 

[Torie]

I do seem to get under your skin don't I Train? I don't mean too. Anyway, I agree that my black CD is more likely than not, not required. I just think is has a better shot then your CD taking in non-contiguous blacks in Birmingham and combining them with rural blacks elsewhere. Some cases have actually mentioned that an urban minority is not part of a community of interest with their far flung rural counterparts, and that term COI is sometimes mentioned in the context of geographic compactness.  But those cases didn't really deal with a set of reasonably contiguous blacks (or other minority). Anyway, let's just add this to the long list of things we disagree about and move on. A black CD is either required in Alabama or it is not, and we will probably never find out (except by analogy via litigation in another state), since both parties want a black CD in Alabama, and I doubt that state will be changing its law to conform to anything remotely like what we are talking about here either.

[muon2]

I believe it quite unlikely Train's black Alabama CD is required by the VRA. To get to Birmingham is a bridge too far. If it is not a bridge too far, than a fortiori, Muon2's NO to BR affair is also required by the VRA. I don't think so.



Perhaps a closer case is the black Bama CD below (it's 49.9& BVAP, but with a change of a precinct's boundaries in Tuscaloosa, can be pushed up to 50%). It has a reasonably contiguous black population, is not hideously erose, but does wander, requires strategic chops, including of a municipality in Autuaga County and the stab grab of a chopped Tuscaloosa municipality), and makes a hash out of the two CD's to the south. (It also chops the Montgomery urban cluster, but I don't think the VRA cares too much per se about urban clusters, although it might take cognizance of them if the state does up to a point.) That CD would make an interesting case, but if I had to bet, it is more likely than not, not required either.

I would draw the same conclusion that AL and LA are in the same situation, and with a need to link two separate urban or urban and rural minority populations, so would VA, SC and NC. I think that if one of these southern states offered a plan with no BVAP-majority CD the court would have to go back to section 2 itself.

You must be logged in to read this quote.

A set of state rules that precludes the creation of an otherwise lawful district, when the potential district uses generally recognized redistricting principles and is designed to give minorities the opportunity to elect their representative of choice, could well be found to have discriminatory effect which is the standard. One measure of discriminatory effect would be that no districts in a state with a significant minority population provided the opportunity to elect representatives of choice for that population. In such a case the rules themselves become unconstitutional.

[traininthedistance]

I do seem to get under your skin don't I Train? I don't mean too. Anyway, I agree that my black CD is more likely than not, not required. I just think is has a better shot then your CD taking in non-contiguous blacks in Birmingham and combining them with rural blacks elsewhere. Some cases have actually mentioned that an urban minority is not part of a community of interest with their far flung rural counterparts, and that term COI is sometimes mentioned in the context of geographic compactness.  But those cases didn't really deal with a set of reasonably contiguous blacks (or other minority). Anyway, let's just add this to the long list of things we disagree about and move on. A black CD is either required in Alabama or it is not, and we will probably never find out (except by analogy via litigation in another state), since both parties want a black CD in Alabama, and I doubt that state will be changing its law to conform to anything remotely like what we are talking about here either.

That's certainly likely; I do think the case for the all-Black Belt district is airtight in a way that the Birmingham-black belt district is, perhaps, not.

Though it must be pointed out that the black belt district includes an urban area anyway- Montgomery.  Granted, Montgomery is a smaller area, and is embedded within the black belt to boot.

[traininthedistance]

So, here's Alabama with the Montgomery-Black Belt district, and otherwise hewing as exactly as possible to the urban county clusters:



And, here's a modification that I'm sure will meet with Torie's approval, where we split up the two counties of the Huntsville urban cluster and, by doing so, drastically reduce erosity in the north:



The latter map is a good illustration of the tradeoffs involved with cluster cuts, and perhaps a good example of why (I agree with Torie on this point) they probably should merely be penalized rather than banned.

In both scenarios, the black district is 50.6% BVAP, with three more-or-less neat county cuts along the perimeter.  Obviously, when drawing VRA districts, one must be prepared to override the usual no-cut stance to accomodate a contiguous minority population which does not hew exactly to the county lines.  In addition, something must be cut in the Birmingham area: while Jefferson-Bibb would make a district, Bibb is not part of the cluster (it is an outlying, rural metro county), and there is no good road connection between the two anyway.  The line I chose minimizes erosity.

For the record, the central Birmingham district (Walker + 95% of Jefferson) is 38.3% BVAP and 52.3% Obama.

[jimrtex]

I see the division of multi-county regions as a separate step, and it can be controlled by limiting the number of cuts to the minimum necessary (N-1), disallowing multi-spanning where two or more districts have parts of the same pair of counties, and emphasizing respect for local communities of interest, particularly where there are local political boundaries available.

Note that I specifically permit forming whole county regions within UCC.  That actually eliminates a (significant) county cut.

In New Orleans, Jefferson+Orleans (1.026), Jefferson+St. Tammany+St Charles+St John the Baptist (1.013), Jefferson+St.Tammany+St.Charles+Plaquemines (0.980), and Jefferson+St.Tammany+Plaquemines+St.Bernard (0.958) are all within a 5% limit.  However, they both  cut off the remnant of the UCC cluster from the rest of the state.

There are cases, to be clear, where local sentiment/communities of interest will in fact prefer multi-spanning of districts.  The canonical example is on Long Island, where the new (and widely praised) court map has two Nassau-Suffolk districts, one each for the North Shore and the South Shore.  IIRC the geography of Washington State seems to suggest that double spanning would be appropriate there as well, perhaps to allow an eastern King district to spill over to Snohomish and/or Pierce, as well as one in the denser western edge. 

It probably would not be that hard to push NY-2 out into Suffolk, and keep it mostly on the south shore.  NY-3 would move more into Nassau and somewhat south, as NY-4 would then be the whole southern part of Nassau.

The court-drawn map has a pretty good adherence to counties, especially compared to previous maps.  But I don't think you can really get that sort of balance between strict compliance and common sense without an "expert" or "judge" imposing a map.

King County ended up being overrepresented in the Washington map.  If you leave out trying to protect incumbents, you would get a reasonable map.

I can understand why one would want a rule against it anyway, but we should recognize that it is not completely without its drawbacks.

And of course I absolutely agree that trying to hew to local political boundaries should be a priority in urban areas.

In large cities, there is no reason that neighborhoods could not be identified well in advance of redisttricting (let's say 2017-2018).  This will ensure that the census bureau can gather the information.  Currently, most of the "evidence" is totally self-serving.

[Torie]

You think SCOTUS would revise its interpretation of Section 2 Muon2, and revise the first prong of the Gingles test, to require a black CD if the percentage population of blacks in that state is high enough, even if not geographically compact?  If so, I don't agree, but in the meantime I would prefer to rely on existing precedents as extrapolated the way a lawyer would extrapolate them to fill in the gaps.

Trains' black CD, or mine, in Alabama, which do have reasonably contiguous populations, is probably close to a 50-50 proposition as to whether section 2 requires a stab into Tuscaloosa that way, to create the 50% BVAP CD (without which none at all is required).  Without the Tuscaloosa stab, I think it more likely than not that the CD's, if 50%% BVAP without the stab, would be required. Sucking up a whole urban county out of an urban cluster (here Montgomery County), sitting within a minority zone, even though the blacks are strung out over a long distance, and the rural blacks might be different than the urban ones, I don't think would deflect a court from that holding in and of itself. And that is particularly true where the CD is not that erose, although it does make another CD rather erose, but erosity collateral damage is probably not going to deflect a court either within limits, which limits are probably hewed to here. 

[jimrtex]

BTW, when did Iberville go back into the BR cluster? I thought that was resolved earlier in this thread and it did not meet the 50% UA threshold.

Note: Terms that are in bold mean the terms as defined by the census bureau.

As an extension of my effort to document which counties belonged and which did not, I changed the criteria to:

25,000 persons in urbanized areas or 25% of the population of the county in urbanized areas.

Since metropolitan statistical areas are based on a core of urbanized areas, I thought it more appropriate to use a test based on urbanized areas rather than urban areas.   Further, the test for being the central county of a core-based statistical area seems to me to be based on using a single definition for both micropolitan statistical areas and metropolitan statistical areas.

That is, since the core urban cluster of a micropolitan statistical area only needs 10,000 persons, it may be reasonable to only require 5,000 in a county to qualify as a core.  But since an urbanized area has 50,000 persons, it seems reasonable to me require more persons within the urbanized area portion of a county, to really regard the urbanized area as being significant.

Iberville is an edge case.  I wouldn't characterize it as hard case.  But it a "county" that is not clearly in or clearly out.

The eastern tip of Iberville is on the east bank of the Mississippi, south of Baton Rouge.  Much of the western part of the county is in the Atchafalaya flood plain used for Mississippi overflow, and is uninhabitable.  I-10 crosses this area on an 18-mile long bridge.  As a result, Iberville does not have much population (33,000) but 11,000 is in the New Orleans Urbanized Area, some on the east bank, and some on the west bank where the urbanized area follows the Louisana Highway 1.  Urban areas in southern Louisiana tend to be strung out along areas protected by levees, compare to places like Monroe and Shreveport, which have a more circluar clumped appearance.

At some time, I will produce a report showing the counties that are included in urban county clusters, which will highlight the counties which are edge cases.

[jimrtex]

So the argument is that UCC's have more of a community of interest than an array of rural counties, and therefore an excess UCC chop should be banned as opposed to merely penalized? I am not comfortable with that argument. If an excess chop makes for a far better map overall, with less chops and erosity elsewhere, it should be permitted. I am not persuaded such a ban is necessary to make for better maps, and less gaming.

Certainly more of a community of interest than a random array of rural counties.  The rule is also intended to protect more rural areas from being placed in districts dominated by urban areas.  When you "chop" cities you are going to be placing more rural areas in the urban-dominated districts.

[jimrtex]

So Jimtex, to cut to the chase, the census bureau identifies the urban percentages in counties, and you excise those that are included in a census urban cluster, but have less than 50% of the population as urban. Is that correct?

Core based statistical areas (CBSAs) and combined statistical areas (CSAs) 

2010 Urban Area to County Relationship File [TXT]

Explanation of the 2010 Urban Area to County Relationship File

The first file defines the counties in each core-based statistical area.  We are interest in metropolitan statistical areas - not micropolitan statistical areas, metropolitan divisions, nor consolidated statistical areas.  That can be disregarded.

The second file defines the intersection of urban areas (urbanized areas are urban areas with 50,000 or more population, urban clusters are urban areas with less than 50,000 population) and counties.

Download the two files and combine them such that you have for each county, its urban areas and metropolitan statistical area.

[jimrtex]

A set of state rules that precludes the creation of an otherwise lawful district, when the potential district uses generally recognized redistricting principles and is designed to give minorities the opportunity to elect their representative of choice, could well be found to have discriminatory effect which is the standard. One measure of discriminatory effect would be that no districts in a state with a significant minority population provided the opportunity to elect representatives of choice for that population. In such a case the rules themselves become unconstitutional.

Our rules are antithetical to the creation of race-based congressional districts.   Residential segregation does not follow county lines.  Alternatively, the methods of producing Section 2 congressional districts are antithetical to good districting practices.

Increasing residential segregation is not a plausible solution.

If one truly wanted the black "community" to have the opportunity to elect candidate(s) of choice, you would provide for segregated electorates, as is done in New Zealand, and some European countries.  But that is unlikely to be accepted in the USA, even though it would be the best way to achieve the stated objective.

So we are stuck with delineating communities based on race while pretending that we aren't delineating communities based on race.   It is not that 99% of the precinct is black, but rather 99% of the voters in the precinct vote for the black-preferred candidate.

You simply are not going to magically fit Section 2 districts into our procedure.

So we need a separate way to identify and draw Section 2 districts, and then draw regular districts around them.

One possible method: Draw smaller districts and then combine them.  While there are not compact areas of black concentration with 750,000 persons in them, there are smaller ones.

So draw a set of smaller districts, with around 8 to 12 times as the number of congressional districts.  Perhaps try 8, 10, 12, as well as 5 and 20 to determine whether these are too extreme.

Combine N of these smaller districts into a Section 2 district.   In the real world, we could let the voters who reside in these smaller districts decide whether they want to have their smaller area combined in such a way.

Remove these Section 2 districts, and apply our process to the residual areas.

Process for creating smaller districts (based on 10 smaller districts per congressional district.

(1) Apportion N = NCD x 10 districts among counties, where NCD is the number of congressional districts.  
(a) Quota = State_Population / N
(b) Ideal Apportionment = County_Population / Quota
(c) Whole Apportionment = floor(Ideal Apportionment).  May be zero for small counties.
(d) Surplus = Ideal Apportionment - Whole Apportionment.

(2) Form apportionment regions of whole counties, whose population is approximately equal to to a whole number of districts.

Maximum error for a region apportioned approximately M districts:

max ( min(5% * M, 7% * sqrt(M)), 20%)

Ideally, the cumulative surplus for an apportionment region is approximately zero or one, which will minimize the number of districts crossing county lines.  It is also preferred to have single-county multi-district regions, or single-district multi-county regions, but this may not always be possible.  Also the total number of districts apportioned must equal N.   This will ordinarily occur if error limits are observed.

(3) Choose best apportionment plan:

(a) Fewest regions with cumulative surplus of 1.5 or more.
(b) Most regions.
(c) Other criteria.   Erosity is pretty useless at this scale.  Popular opinion may be considered.

(4) Subdivide apportionment regions.

(5) Determine if districts may be organized into a congressional-district sized unit.  I would not require contiguity.  The community of interest does not require contiguity in order for it to be represented.  It is possible that popular opinion could be considered.  Your right to vote is not being protected if race-based districts are being imposed on you.

(6) The overall error for the congressional district may be a bit outside the 0.5% limit.  It would probably be easiest to make the correction after the other regions are defined.

(7) Remove population of Section 2 districts, and create remaining regions.

[jimrtex]

Returning to the ban versus penalty issue for urban cluster chops, to me that issue turns in the end on high the likelihood is that it would generate ugly maps. I view it as unwise to go there unless and until that specter is put to bed. Univariate rules tend to worry me, when what we are doing in the end is all about finding the Golden Mean, balancing off this against that. I am also unsure how wise a ban would be on policy grounds as well (putting urban clusters and their separation from the rurals, and visa versa, at the top of the heap of considerations, as opposed to but one consideration among others). The rationale is that it further helps to rein in gaming seems weak to me. A penalty would do about the same thing I would think. In the end, the score is the thing.

OK.  Double the erosity for all districts that chop an UCC.

[muon2]

A set of state rules that precludes the creation of an otherwise lawful district, when the potential district uses generally recognized redistricting principles and is designed to give minorities the opportunity to elect their representative of choice, could well be found to have discriminatory effect which is the standard. One measure of discriminatory effect would be that no districts in a state with a significant minority population provided the opportunity to elect representatives of choice for that population. In such a case the rules themselves become unconstitutional.

Our rules are antithetical to the creation of race-based congressional districts.   Residential segregation does not follow county lines.  Alternatively, the methods of producing Section 2 congressional districts are antithetical to good districting practices.

Increasing residential segregation is not a plausible solution.

If one truly wanted the black "community" to have the opportunity to elect candidate(s) of choice, you would provide for segregated electorates, as is done in New Zealand, and some European countries.  But that is unlikely to be accepted in the USA, even though it would be the best way to achieve the stated objective.

So we are stuck with delineating communities based on race while pretending that we aren't delineating communities based on race.   It is not that 99% of the precinct is black, but rather 99% of the voters in the precinct vote for the black-preferred candidate.

You simply are not going to magically fit Section 2 districts into our procedure.

So we need a separate way to identify and draw Section 2 districts, and then draw regular districts around them.

One possible method: Draw smaller districts and then combine them.  While there are not compact areas of black concentration with 750,000 persons in them, there are smaller ones.

So draw a set of smaller districts, with around 8 to 12 times as the number of congressional districts.  Perhaps try 8, 10, 12, as well as 5 and 20 to determine whether these are too extreme.

Combine N of these smaller districts into a Section 2 district.   In the real world, we could let the voters who reside in these smaller districts decide whether they want to have their smaller area combined in such a way.

Remove these Section 2 districts, and apply our process to the residual areas.

Process for creating smaller districts (based on 10 smaller districts per congressional district.

(1) Apportion N = NCD x 10 districts among counties, where NCD is the number of congressional districts.  
(a) Quota = State_Population / N
(b) Ideal Apportionment = County_Population / Quota
(c) Whole Apportionment = floor(Ideal Apportionment).  May be zero for small counties.
(d) Surplus = Ideal Apportionment - Whole Apportionment.

(2) Form apportionment regions of whole counties, whose population is approximately equal to to a whole number of districts.

Maximum error for a region apportioned approximately M districts:

max ( min(5% * M, 7% * sqrt(M)), 20%)

Ideally, the cumulative surplus for an apportionment region is approximately zero or one, which will minimize the number of districts crossing county lines.  It is also preferred to have single-county multi-district regions, or single-district multi-county regions, but this may not always be possible.  Also the total number of districts apportioned must equal N.   This will ordinarily occur if error limits are observed.

(3) Choose best apportionment plan:

(a) Fewest regions with cumulative surplus of 1.5 or more.
(b) Most regions.
(c) Other criteria.   Erosity is pretty useless at this scale.  Popular opinion may be considered.

(4) Subdivide apportionment regions.

(5) Determine if districts may be organized into a congressional-district sized unit.  I would not require contiguity.  The community of interest does not require contiguity in order for it to be represented.  It is possible that popular opinion could be considered.  Your right to vote is not being protected if race-based districts are being imposed on you.

(6) The overall error for the congressional district may be a bit outside the 0.5% limit.  It would probably be easiest to make the correction after the other regions are defined.

(7) Remove population of Section 2 districts, and create remaining regions.

There seems to me an easier way to accomplish much the same result. Much of this problem seems to arise from multi-county areas that have a more dispersed minority population than is found in urban centers. Often these are rural counties, but could also be counties with smaller cities.

One way to address this within the existing framework is to identify counties that have significant minority populations and could contribute to a section 2 district much like we have identified urban counties. Contiguous clusters of such minority-contributing counties could function in much the same way as UCCs. Like a UCC they would be embedded in a region, and embedded in a way which would enable the region to support a section 2 district. A region could be comprised of both whole UCCs and MCCs (minority county clusters) much like a region can have multiple UCCs. As with the current concept regions would still be required to be nearly a whole number of districts and more regions are preferred in a plan (subject to erosity).

Then within a section 2 region one would subdivide counties into smaller units to use for the formation of districts. Rather than use an artificial sized unit, one should use municipalities or other Census-based county subdivisions. That would be consistent with our sense of how chops should be made. Erosity would a factor when choices are available for the formation of the districts, but county chops within a section 2 region would not matter as much as county subdivision chops.

[Sol]

In a similar vein to the discussion of Louisiana-Is it justified to split the Tucson Urban County Cluster to create a Hispanic district?

[muon2]

AL has a well defined black belt so, I've put together a map of the counties to see how a minority cluster might be constructed.



The key to the colors is:
yellow 25.0-33.3% BVAP
lime 33.4-39.9% BVAP
green 40.0-49.9% BVAP
dark green > 50.0% BVAP

At a minimum the dark green counties would have to count in the cluster. A threshold of 40% doesn't seem unreasonable either, they are like the outlying counties in a UCC that have enough UA population to count in the cluster. Thresholds at 1/3 or 1/4 start to look too extensive beyond the essential idea of a cluster.

The AL black belt cluster overlaps the Montgomery urban cluster. My thought is that a region should not split the UCC, but a district might if satisfying section 2. One thing I observed in 2011 was that at times there was determination that a section 2 district was required, but beyond that the only guidance was that the district should be able to elect the candidate of choice of the minority. This resulted in districts that were not necessarily covering all the areas that gave rise to the requirement, but yet resulted in a qualifying district. For instance a black neighborhood in St Clair County IL gets sliced out of a compact VRA district and added to a neighboring district for political purposes, yet the remaining district can be linked to more remote crossover voters to get the minority candidate of choice elected. Perhaps this gets to Torie's observation that a particular district is not required, yet I can look at the whole plan and believe the VRA requires a district somewhere as long as race is not the sole factor and the district is not a "geographic monstrosity."

[Torie]

I have never had any problem with the notion that the VRA will next to insist on creating a majority-minority CD which otherwise takes in a reasonably compact and contiguous population, that involves shaving off a whole county from an urban cluster. That will probably happen. It is when it comes to more chops, non contiguous and more erose, that the field becomes more opaque, until at some point I next to not see it at all.

So on this one, to the extent of the above, and also that that 40% figure of yours looks like a reasonably horizon spot to seek out such minority clusters (at least to the extent reasonably contiguous, and not too erose), we are on the same page -  for once! 

[jimrtex]

I have never had any problem with the notion that the VRA will next to insist on creating a majority-minority CD which otherwise takes in a reasonably compact and contiguous population, that involves shaving off a whole county from an urban cluster. That will probably happen. It is when it comes to more chops, non contiguous and more erose, that the field becomes more opaque, until at some point I next to not see it at all.

So on this one, to the extent of the above, and also that that 40% figure of yours looks like a reasonably horizon spot to seek out such minority clusters (at least to the extent reasonably contiguous, and not too erose), we are on the same page -  for once! 

I'm not familiar with use of "next" as a verb?