Local vs regional road connections

[jimrtex]

There are paved roads between the two counties, just not state highways. Don't those count? I thought they did. I believe that they should, even if not for erosity purposes. I don't see the policy reason why not, given the further limitation on flexibility, and potentially forcing chops elsewhere. Yes, the alternative is to chop into Mecklenberg, but then that chops into Charlotte, which is the same as a county chop.

US 52 provides a direct connection between Salisbury and Albermarle. That it nicks the extreme north-eastern corner of Cabarrus is of no consequence. Rowan and Stanley share the Great Pee Dee River as their eastern boundary.

If the classification would occur at the time the map was drawn, it could possibly be used to gerrymander.

If the classification were established well in advance of the map drawing process, there would be no reason to restrict such direct transportation connections.

[jimrtex]

There are paved roads between the two counties, just not state highways. Don't those count? I thought they did. I believe that they should, even if not for erosity purposes. I don't see the policy reason why not, given the further limitation on flexibility, and potentially forcing chops elsewhere. Yes, the alternative is to chop into Mecklenberg, but then that chops into Charlotte, which is the same as a county chop.

Whole counties have to be regionally connected, ie by means of all-season numbered state/federal highways or ferries. The policy is useful to eliminate weaker local connections across natural barriers like mountains, rivers and deserts and as a proxy for the relative amount of contiguity two counties have. The only exception we've made in the past is for counties that are in the same UCC, where local connections are sufficient. The alternative for flexibility would be to assess additional chop penalties to discourage inter-county linkage based only on local connections.

I am not sure I agree with that. The compromise, which is reasonable, is no erosity issue without state highway cuts, but no prohibition of links either if paved. It seems a bit too arbitrary, and reduces flexibility, which is needed here to hew to the urban cluster rules. This metric would force a deviation. I think respecting urban clusters is more important myself.

The problem is that the rule doesn't "know" about whether it's being used to protect a UCC pack. The UCC rules should be no more important than other rules (and based on the voting by Atlas members last year, less important). If connectivity is changed to allow local connections and not just regional connections then it affects all counties everywhere, and we can have some unpleasant shapes that simply take advantage of the rule. For sure you can get cross-mountain districts in the West, since that's an area that was used to develop the rule in the first place.

And yes I know about US 52 cutting a corner of Cabarrus. The metric is whether you can go between two county seats on state highways without going into a third county. It is easy to apply. The moment the door is opened to allow a road to go a little bit into another county the metric becomes exceedingly messy as one tries to define how much goes beyond a little.

The rule lacks in common sense.

[jimrtex]

Are you now ok with the CA county connections that I mentioned? As I said the rule arose in discussion about mountains out west, where such connections were not seen as good policy. The rule was tested in Midwestern states (MI in particular) with counties that had minimal overlaps of boundaries, no highway connections, but nonetheless had local roads.

California did not consistently apply the rule about coming into San Benito County via a road that sneaks into Santa Clara County. It appears that those who advocated for the separation were doing so for political expediency.

[jimrtex]

And yes I know about US 52 cutting a corner of Cabarrus. The metric is whether you can go between two county seats on state highways without going into a third county. It is easy to apply. The moment the door is opened to allow a road to go a little bit into another county the metric becomes exceedingly messy as one tries to define how much goes beyond a little.

The rule lacks in common sense.

When I spoke in 2010 with redistricting experts who had mapped states in previous cycles they said that one of the easiest ways to gerrymander is to grab populations across a river or forest that you couldn't otherwise conveniently reach, but were contiguous. One even suggested that if he was to only propose only one change to reduce gerrymandering it would be that it had to be possible to reach all parts of a district by car without leaving the district. That sounds like common sense to me.

If you were to put Rowan and Stanley in the same district, you would not be crossing a river or forest. You would be nicking the corner of Cabarrus County.

Do you think that county officials in Rowan and Stanley never work together because Cabarrus County is "between" them? If their were an auto accident on US 52, who could get there fastest - someone from the Rowan or Stanley sheriff's office, or someone from Cabarrus County?

While one person suggested that you must be able to drive to all parts of a district, he probably had done a superficial analysis of what that literally means.

[jimrtex]

This is a network representation of North Carolina counties superimposed over an outline map of the state.



The nodes (blue circles) represent the counties, including the boundaries, population, etc. The position of the node over the map doesn't provide any information. I've used the internal points of the counties calculated by the census bureau.

The links represent the contiguity relationship between counties. Two county nodes are linked if the counties they represent are contiguous.

The red squares represent point contiguity, which we are not permitting. If we did, then links would cross. This would not make the network invalid, though it would make it non-planar. We don't permit use of point contiguity for policy reasons. Districts that cross each other are decidedly non-compact, and are little better than non-contiguous districts.

In building a district from whole counties, we can imagine snipping the links between counties in different districts. Rather than snipping the center of the link, and leaving the remnants dangling, we completely remove the link.

Alternatively, we can build a district by traveling along the links collecting counties and population as we go.

The more links we snip, the less compact our district will tend to be. I suspect it is possible to create a district from the western tip, northeast along the Tennessee boundary, across the state along Virginia, and then down the coast to Wilmington. We would snip a large number of links, or alternatively retain very few as we create a string of 30 counties.

The concept of connected counties is that some counties have a stronger relationship than mere contiguity. It is a COI concept, suggesting that they are neighbors, rather than merely contiguous. The central idea is that you can easily travel between the counties. The roads were created because there was a need to travel between counties, the existence of roads then reinforces the relationship between the counties.

In some cases it is difficult to travel between adjacent counties. There might be water or mountain barriers. The contiguity might be limited and no easy way to travel between the counties.

We can demonstrate the concept with states. Kentucky and Missouri are contiguous, as are New York and Rhode Island, and Minnesota and Michigan. It is not easy to travel between the pairs of states and would likely require a boat.

If we were traveling between New York City and Providence, Louisville and St.Louis, or Detroit and Minneapolis, we would almost assuredly not travel directly from one state to the other.

Traveling from New York City to Providence we would travel the length of Connecticut. From Albany we would likely travel across most of Massachusetts. New York and Rhode Island are not connected.

Driving from Detroit to Minneapolis, we would likely drive through Chicago. We could use the ferry across Lake Michigan, but not in winter, and we would still have to cross Wisconsin. We could travel through the Upper Peninsula and drive quickly through Superior, but the route is circuitous. Michigan and Minnesota are not connected.

You can drive from Louisville to St. Louis in a direct route. If you drive across the southern tips of Indiana and Illinois you avoid most of the people. And you can take a somewhat longer route across Kentucky and then drive into Missouri.

But this route requires a crossing into Illinois south of Cairo. Even though less than a mile is within Illinois, Muon2 would argue that the two states are not connected. I believe that they are.

On the network, the links between counties that are clearly connected are in blue. Those that there are contiguous but possibly not connected are in yellow. Muon2 and I likely agree that some are not connected, and perhaps that some are connected.

[jimrtex]

You must be logged in to read this quote.

Fixed.

Well said jimrtex, especially the CoI explanation.

Our differing views stem from my strong propensity to avoid paths that cross outside the district. In this case that means paths that go outside the two counties considered for a connection. I think Torie favors this path containment as well, but wouldn't impose the CoI standard of a state highway.

I think that makes four categories of links.

1) Contiguous: the areas share a border greater than a single point.
2) Locally connected: contiguous areas where public roads provide a link without passing through an intervening county.
3) Nearly connected: contiguous areas where there is a convenient state highway linking them, but the path may briefly pass through intervening counties.
4) Regionally connected: contiguous areas where state highways provide a link without passing through an intervening county.

I view the approach as building a region of a whole number of districts by following connections. Then separating regions of more than one district using a minimum number of chops. The FL Senate exercise was a case in point.

My viewpoint is that each node represents a sub-network of areas within the county. We collapse the subnetworks into a single point for simplification, and to strongly stress the use of whole counties. We only need to examine these subnetworks, if we are forced into splitting counties.

In Florida, I approached it as a matter of decomposing the statewide problem into smaller more manageable problems. After dividing the map into apportionment regions, the division into individual districts can be handled independently. In Florida, there could be redistricting commissions for each county, that would delineate commissioner districts, and possibly city council districts. If necessary, they would participate in the division of their counties into districts.

In Florida, there were a couple of different ways that the regions could be arranged. At that point, the process could be forked, and then the decision on the statewide map could be based on which produced better districts in the alternative regions.

Chops modify the connection network.

Chops less than a specified size (5% for CDs) split a single node into two connected nodes. The existing connections are assigned to the respective nodes based on where the primary path associated with the connection crosses the border.

If all chops are treated as above, districts in populous urban areas come out with artificially low erosities regardless of their shape. Chops in these areas need a denser network to assess district shapes and maintain parity with large multicounty rural districts.

So, chops in excess of the specified size are macrochops. Macrochops split the county into a network of nodes based on agreed subunits. Connections are made between these new subunits and from subunits to adjacent counties. Based on the Kent exercise, subunits use local connections within a county and regional connections to adjacent counties. The limitation of cross-county connections to regional links for subunits was an important result from the work in MI to maintain the rural-urban balance for erosity.

If a wider deviation range is permitted, you can do away with some chops. Chops needed to equalize between regions can be treated as independent problems. Chops within multidistrict regions are not chops at all. Respect for counties is a constraint on the division of townships, cities, and other subcounty areas.

[jimrtex]

I'm giving your model a road test as it were. You should thank me, rather than be annoyed.

Great then we should see how well it performs under the current rules. If it fails to produce a reasonable map in multiple occasions, then it should be modified. One outlier isn't convincing.

Well I have already pointed out some of the downsides based on actual mapping experience. I guess in the end, states if they have any interest in your approach, will make up their own minds. I still think that using the blue roads for erosity testing, and allowing the yellow roads per jimrtex's NC map, for CD's to make use of, is a reasonable and sensible compromise. I would still like to see what you were talking about with Kent. I thought that was about internal connections, not cross county connections.

You misunderstand my map.

The links are not roads. They are a graphical representation of a relationship between counties, that of contiguity. The network is superimposed over the map to illustrate the derivation of the map.

The links can have attributes associated with them. For example, they could have the length of the border between the two counties. Certainly, they could indicate whether the two counties are "neighbors" or merely adjacent.

Some of the yellow links could become red links, indicating they are not neighbors, even though they are adjacent. Others could become green, indicating they are neighbors.

The blue links would all become green links. The blue links are those that are trivially obvious. It is easy to travel directly between the counties, using numbered highways.

Muon2 and I would disagree on the further classification of some of the yellow links, but would likely agree on others.

The census bureau has a county adjacency file for the entire country, which shows the contiguous counties for each county.

I extracted the North Carolina portion; and removed the interstate adjacent counties (to Virginia, Tennessee, Georgia, and South Carolina). The census bureau classifies each  county as being adjacent to itself, so the self-adjacent were eliminated. And the census bureau lists counties adjacent to each county, so that Kings (NY) is adjacent to Queens (NY), and Queens (NY) is adjacent to Kings (NY). I removed the second copy of these duplicate links.

This left 256 adjacent pairs of counties. Another 12 were eliminate due to point contiguity. For example, Guilford and Stokes are adjacent as are Forsyth and Rockingham. This could be automated. I did it with a spreadsheet, though I manually entered formulas, and stepped through the process.

It might be possible to automate further classification of the links, but I did it by hand. It can certainly be assisted by computers. It could also be farmed out to the counties. If two counties agreed, it would probably be accepted by the state.

I would not permit use of red links for building districts. I probably would use them for measuring erosity. Generally, they correspond to fairly short boundaries, so that there is little penalty.

[jimrtex]

I've been doing some research on the NC road system. They claim to have the greatest number of road miles in the US. That's in large part because there are no county roads in NC. Any road not maintained by a municipality is maintained by the state. Most of those roads are called secondary roads (SR) and some have numbers used for DOT purposes, but they aren't always marked and they aren't intended for navigation. They move them up to primary state roads when they are of regional significance.

For instance the roads Torie initially posted on the Montgomery-Davidson border are SR's. Mapquest doesn't show those numbers at all and Bing maps only show a number on the Montgomery side (Blaine Rd  NC - 1161) but not on the Davidson side (Badin Lake Rd). Based on the NCDOT guidance I would classify them as local roads rather than numbered highways.

OTOH, there are lots of legitimate numbered state highways that aren't "highways". I'm going through all the yellow links on jimrtex's map to see how I would classify them. I'll post that soon.

I used the primary and secondary roads from the census bureau, along with Google Maps.

I think that they are largely provided by the counties. In other states they are sometimes inconsistent, with some counties making a greater effort. There is likely some bias in classification - but that is not necessarily a bad thing. In older areas of the country, county boundaries are more likely to reflect settlement patterns,  with roads built to connect the somewhat isolated towns. Towns would develop as market places, and to provide for the little bit of government needed by a rural population.

When roads were improved, one goal would be to make it easier for farm wagons to get to the market town. Later, these would be paved so farm trucks could do the same, and then so rural dwellers could get to work in mills. As such most of the road systems look like hub and spoke, with the road system in each county town looking like a mini-Boston or micro-Chicago.

That the roads look like they were purposefully reaching out to all corners of the county, and directly to neighboring counties, is likely because they were purposefully built for that purpose.

Factors that I used in classification were:

Adjacency Index

Length of the border gap divided by the square root of the area of the smaller county. If two midwestern counties were 5 (survey) townships square, and they overlap by one township, the adjacency index is 0.20.  At that point, the physical connection is somewhat weak. In Iowa, with its primarily east-west and north-south road system, there might not be road connections along diagonals. To travel between the counties, you travel 24 or 30 miles north, and then make a right turn and travel 30 or 36 miles west. While you might be able to zigzag along section-line roads, you probably wouldn't. If there was a reason for significant northwest, etc. travel such as for commuting, the roads would be improved, the counties connected.

When the adjacency index is below 0.10 the contiguity may be almost accidental, provided by a panhandle, or foot, or some other extension. People might not really think of the county as a neighbor.

Someone in Warren County, might think of Granville and Northampton as both being moderately near, but not close, even though Warren and Northampton are adjacent.

Direct Road Connection

This would probably be best measured using trips between residential locations (perhaps using a random selection of pairs). Possibly better would be using commuting targets, but that is harder to weigh. Direct means traveling in a purposeful way, not circuitous, or via other places.

Extent of Travel in Intervening Counties

Traveling from New York City to Providence, the travel through Connecticut is not an inconsequential part of the trip. Traveling from Louisville to St.Louis using the bridges south of Cairo, Illinois is not a significant portion of the trip in either time or mileage.

The extent could also include how much of the population of the intervening county is along the route. If you are traveling on US 52 southeast from Salisbury, NC, you are unlikely going to a location in Cabarrus County.

Water Contiguity

In general these require a circuitous route, but a ferry is an acceptable alternative.

[jimrtex]

"OTOH, there are lots of legitimate numbered state highways that aren't "highways"." 

What does that mean?

It means they didn't show up on jimrtex's map as blue links, but I would include them.



In this map I have revisited the yellow links on jimrtex's map.

Blue are regional connections based on continuous state highways between two county seats that don't enter a third county. The dark blue links were those not identified previously, but meet my criteria.

Green is a all year ferry connection that meets the criteria for a regional connection.

Yellow are local connections that rely on local roads to establish a path between counties.

Orange are near connections based on state highways where the highway path cuts a short distance through a third county, such as at a corner.

Gold are connections equivalent to both yellow and orange.

Pink are contiguous counties without a connection.

Red squares are places with point contiguity.

This is my interpretation. Note that Haywood and Henderson are actually more than point contiguous.



These are the differences. I have counted everything that Muon2 classified as pink, yellow, gold, and orange, as false.



The 9 instances where I considered the counties to be connected, but Muon2 does not, appear to be cases where the counties had significant boundary connectivity and a direct road connection, that happened to nick the corner of another county, or go along an edge, without going near its center.

The 8 instances where Muon2 considers the counties to be connected, but I don't, would generally bypass other populations.

[jimrtex]

That all seems impossibly complicated to administer. And most of it seems more about erosity as what is in issue, ran than a flat out ban on having the two counties in one CD if accessible to each other by a paved road. I suppose there is some merit however as to whether the road is a direct one to the county seat, or directly to a road which connects to the county seat where there are mostly north south roads, so you need to make a turn at some point. That yellow road between Staley and Rowan was circuitous, although one of the white paved roads I think was more direct. The point I guess is whether the paved road heads to some population center in the adjacent county, as opposed to just wandering around farm areas, services the farms and the scattered residences, and is really not used for inter county travel. But again, we are slipping into a more subjective area here. I suppose one could set up the parameters, and then each state decides what roads do what, so they make the more subjective calls.

I don't understand what you are saying.

[jimrtex]

Here is an illustration on how I view the interplay between (macro)chops and connections. I start with the county connections, like the blue lines in the state map below.



When a macrochop (>5% of a CD) occurs, then the single node of the county is replaced by a network of nodes associated with the county subunits. Mecklenburg is a good example to use, since it has to be macrochopped in any plan. NC doesn't have functioning townships in the way MI does, and unincorporated areas are annexed to the nearby towns when the population gets large enough. So I created subunits for the 6 independent towns and the 6 planning areas for Charlotte from the city website. All of the precincts are assigned to one of these 12 subunits as nearly as practicable. The subunits are shaded on the map below.

Each of these subareas is connected other subareas within the county if they are locally connected as shown by the gold links. The subareas are then connected to the adjacent county nodes as either regional connections (blue), local connections (gold), or contiguous but disconnected (pink). These are all shown on the map below as well.

At this point I can proceed as if this was the new state map with a bunch of extra counties (the subunits) where Mecklenburg was. My rules currently treat the gold links inside Mecklenburg as valid connections, but only the blue links that cross county lines as connections. Erosity is calculated the usual way by evaluating which links are severed.



Note that macrochops create new links both within the county and to adjacent counties. If this weren't a macrochop, I would only replace the county with one subunit for each chop fragment. The only new links would be the internal connections between fragments.

Under a decompositional model you can simply treat all county areas as nodes, and divide each apportionment area independently from the state. You would still have a constraint of limiting the division of counties.

[jimrtex]

What does generally bypass other populations mean precisely? I notice that between Haywood and Madison, there is no state highway that gets to the county seat without departing from Madison again.  In fact there appears to be no paved road at all. The state highway just nicks Madison in its corner, and then departs Madison again. I am surprised Muon2 considers that connected. Maybe he is thinking about a chop that just takes in the connected part. Yet you say that if the state highway does not cut in at the "center" of the county, Muon2 rejects it. What is an example of that? That does not seem consistent in any event with Muon2's acceptance of the Haywood-Madison link.

It seems to me that policy point is to have a paved highway that is designed to get from county seat to county seat however it is numbered, even if you have to make a turn on to another highway of some sort. And maybe under that standard, nicks are OK, provided there is some other paved connection that is paved that does not force a departure from the county (the idea that you can travel from county to county within a district without ever leaving the district that that guy Muon2 paraphrased stated). Or perhaps not.

"The 8 instances where Muon2 considers the counties to be connected, but I don't, would generally bypass other populations."

I should have written:

"In general, the 8 instances in which Muon2 considers the counties to be connected, but I do not, the connection would permit other population to be bypassed."

That is, I meant to characterize the differences in our methodologies, rather than characterizing the connection itself. The connection bypasses a population. That is in general the difference in our methodologies, or at least the effect of those methodologies. I may reject a connection if it is not direct. Indirect connections often bypass or go around other areas.

As you know, one gerrymandering technique is to connect two separated concentrations of population while bypassing an intervening population (for example the Chicago-area earmuffs district, where you loop around a population that you don't want to be in the district.

I assume that Muon2 is using NC 209 northeast from Waynesville to Hot Springs, and then south to Marshall. This is a scenic byway that crosses over Betsy Gap and then drops down to Hot Springs. Part of the road is not recommended for commercial trucks, presumably because of the grade and curves.

It is much faster and shorter to go through Asheville. In addition it appears that much of the population of Madison in the southern part of the county, where it is possible to get to Asheville reasonably easy. The largest town in the county is Mars Hills, but that is only about 1600 of 20000 persons.

It really doesn't matter too much, since Haywood and Buncombe are both in the Asheville UCC.

I think that Muon2's methodology is self-consistent in this case, since NC 209 does indeed go from Haywood to Madison, without crossing in to Buncombe. I would reject the use of NC 209 because it is not a direct way to travel between the counties. It is circuitous, and 65% slower than the alternative.

I said that I would accept a direct route that crosses a corner or edge of an intervening county, so long as it avoids most of the population of the intervening county. That is I would treat Rowan and Stanly as being connected by US 52. You would use the route to travel between Salisbury and Albermarle.

The only time Cabarrus would come up is if there is a sign that says "Entering Cabarrus County", and someone commented: "I thought Cabarrus was down by Concord" and the reply, "it has an odd corner up here."

Or Cabarrus were wet, and Rowan and Stanley were dry, so that there were three liquour stores on that one mile stretch of highway.

Or the Rowan and Cabarrus county sheriffs patrolled the highway, but you had to pay your speeding ticket in Concord.

If US 52 crossed more substantially into Cabarrus, or if you went into Cabarrus and switched highways and then drove to the other county, you would no longer be driving directly between the two counties.

I'm not sure whose policy you are referring to in the 2nd paragraph.

[jimrtex]

Would you allow the nick of Cabarrus to suffice if there were no road connections at all between Rowan and Staley?  Would you reject a link that went to the county seat, but made a wiggle to avoid a closer population node in the county?  And you would reject a connection, if there were a more direct connection that went through another county I take it. Is that correct?  How do you define what is too large a population for a nick to work?

You may be overthinking this.

The purpose is to exclude really bad cases of contiguity, rather than picking best connections.

Imagine you were assembling a wall from county-shaped stones. For the mortar to bond, you prefer a longer surface. And you might want to have some reinforcement between the stones such as metal rod drilled into them.

The direct road is a demonstration that the border between the counties is not a barrier between the counties, that people go back and forth easily and regularly. If that is true, then it is OK to put the counties in the same district because they share a community of interest. It will also tend to result in more compact districts.

This is one example of where Muon2 and I differ.



There is a road down the Moore peninsula to Cumberland, but the border is quite narrow. It does not mean that Cumberland and Moore could not be in the same district. It would simply require inclusion of Hoke or Harnett.

Because Fort Bragg is in the middle of this area, there may be a stronger community of interest than indicated by the boundaries. The population is overwhelmingly concentrated in the southern part of the county, Carthage is simply not representative of the population. From the concentration of population around Southern Pines and Pinehurst there are two ways to get to Fayetteville - around the north side of Fort Bragg, or around the south side. It is a quirk that the northern loop is in Moore County.

To your questions.

Whether or not there were other roads doesn't matter. Incidentally, Google suggests a different route than US 52, for traveling between Salisbury and Albermarle. I suspect that there may be a difference in getting out of Salisbury from whatever point they identify as "Salisbury".

I would in general consider where the people actually live. In many counties the road system looks like a spider web radiating out from the county seat. The county seat is the focus of the county. In other counties, that is not true, particularly those that have lots of recent development that has changed the settlement pattern.

Roads typically do not avoid cities. They are created to go where the people live or work or trade. People live in cities.

Consider Henderson, Polk, and Rutherford counties.

The population in Henderson is concentrated around Hendersonville, with some along the northern line around Fletcher, which presumably is more Asheville-oriented. Rutherford is concentrated around Rutherfordtown, and Forest City.

There is a not an entirely unreasonable route that uses US 64 and US 74A. It is not totally direct, since you really are going northwest and then making a right turn and going southeast.

Even though Hendersonville and Rutherfordtown are a bit north of US 74, the route along US 74 is much more effiicent (43 minutes vs 64 minutes) even though the mileage is the same. If you have to drive 38 miles at an average speed of of 36 MPH, you are not going to take that route on a regular basis. The southern route at 53 MPH would be used, and that takes you through the length of Polk.

These three are in mountainous areas where there may be technically available roads, but they are slower or more circuitous because of the terrain.

Haywood-Madison
Henderson-Rutherford
Yancey-McDowell

Hyde County includes Ocracoke Island. There is a ferry from Cartaret to Ocracoke. Most people probably either stop there or go on to Hatteras Island. But there is another ferry north to the mainland and the Hyde County seat of Swan Quarter. Via ferries it is 7 hours from Beaufort to Swan Quarter. By road it is 2:31 hours, and through two other counties.

Carteret-Hyde

I've discussed this one above.

Moore-Cumberland

These involve offset corners in metropolitan areas. and could reasonable be seen either way.

Davidson-Guilford

Most of the population of Davidson is in the north, and it is part of the Winston-Salem UCC. But there would certainly be suburbs along the Davidson-Gulford line oriented more towards High Point (and Greensboro). I was concerned about bypassing Randolph. Since the population is concentrated toward Greensboro and Winston-Salem it might not matter, particularly since Davidson is in the Winston-Salem UCC.

Chatham-Durham

The direct route between Chatham and Durham goes through Chapel Hill in Orange County, which is the largest concentration of population. Since Orange and Durham are in the Durham-Chapel Hill UCC it might not matter.

For our purposes it might be simpler to treat multi-county UCC as a single node, particularly if enforcing a cover rule.

Johnston-Nash

This bothered me a lot since it tends to cut Nash off from Wake (Raleigh), particularly since Johnston is part of the Raleigh UCC. The roads clearly indicate that southeastern Nash is Raleigh suburban. But it also includes Rocky Mount which spills over into Edgecombe County. From Nash to Rocky Mount, you go through Wilson.

BTW, did you know you had a micropolitan statistical area with your name?

[jimrtex]

I've had a long busy day.

I'm not sure what the beef is about NC 209 connecting Haywood and Madison. I put Waynesville and Hot Springs into Mapquest, Bing and Google Maps as end points. All returned NC 209 as the preferred route with the shortest time and the route crosses into no other counties. From there I go down US-25 to Marshall. If I place Marshall as the end point, NC 209 still comes up on Bing as one of the choices, even if not the shortest (1 h 18 min vs 1 h 1 min which is 28% longer). I try to take a consistent approach to connections, without a subjective judgement about their nature in terms of efficiency.

I use Mapquest and Bing as cross checks on each other since both have renderings of the county lines as I magnify the scale. Google Maps which does not show county lines. Bing is also the mapping for DRA so there is some consistency in using Bing, but the DRA version uses a 2010 Navteq file.

This is why I would use local knowledge to make the determination.

Google: 54 min/92 min (70%)
Mapquest: 56 min/87 min (55%)
Bing 61 min/78 min (28%)

Waynesville is about 3000 feet, Betsy Gap is 3901 and Hot Springs is 1375, so there is a big drop coming down into the French Broad valley. NC 209 is designated as a scenic byway, and the portion down from Betsy Gap is not advised for commercial trucks.

The maps agree on mileage (50 miles via Asheville, and 57 via Hot Springs).

But Bing would indicate 49 mph on a route that has much interstate, and 44 mph on a winding mountain road. While Google would have it 56 mph and 37 mph.

I'm very skeptical of Bing in this case.

One aspect of 'direct' is efficiency. A person who is direct, is straightforward, and doesn't spend a lot of time dawdling around the countryside. With a direct flight, you don't have to make connections, and will generally be quicker, and probably a shorter distance.

Mapquest and Bing both render the county lines at Little Pisgah Mt as a four-corner junction. Bing kept it as four-corner to as small a scale as I wanted. That led me to place a red square there.

The census adjacency file does not include Buncombe-Transylvania. When identifying the four corners, I was cross-checking for the alternative, before removing the apparent point contiguity. My GIS program shows a point, as do the shapefiles. But the census bureau does show a small boundary if you use American Factfinder.

USGS Topo Sheet (7.5' Dunsmore Mountain) does show a Haywood-Henderson border of about 150 m.). The Transylvania-Henderson line is a straight line to the peak of Little Pisgah Mountain. The Haywood-Buncombe line appears to be following the ridge line from the west (see 7'5 Cruso), but actually is on the north side of the ridges. I suspect that there was a survey error, with the intent to follow the ridge line, but the coordinates actually being north of the ridge line.

The result is that Haywood-Buncombe line hits the Bumcombe-Henderson just to the north of Transylvania-Henderson.

This is no bunk.

[jimrtex]

The Henderson-Rutherford connection is based on Hendersonville to Rutherfordton (those are easy seats to keep track of ). Bing lists three choices with I-26/NC-108 as the best at 37.3 mi and 49 min (light traffic), that's one that goes through Polk. The third choice follows US 64 at 39.0 mi and 75 min without going through Polk. There's no question that US 64 is a main highway, and it doesn't bypass any significant population center. 64 is one of two US routes that cross the ridge on the eastern border of Henderson, but because it was not improved to a limited access highway it takes 50% longer.

I don't have an objective threshold to say when one route is too much longer than another, and I'm more concerned about keeping paths within a single district, so I count it. That second point is why I disliked VA-3 so much. It relied on a number of contiguous water crossings without a path from one end to the other that stayed in the district. I'm willing to go out of my way to insure a complete path.

If you cross the James River from Newport News, you end up in Suffolk, rather than Portsmouth.  But if you include Suffolk, you can go all the way to the state line. So do you chop Suffolk to get a road connection into Portsmouth?

I had not realized that US 64 turned at that junction and headed southeast. If you follow US 64 across North Carolina, it looks like it was pieced together - likely to secure federal funding.

Edit: Thanks for the info on Little Pisgah. I wonder why the GIS data doesn't show a 150 m border. There are other places where I've seen borders that short.

Differences between 2000 and 2010 maps. When you download shapefiles from the census bureau you get to a page where you can select 2000 or 2010 shapefiles, even though 2010 was previously selected.  2000 is first, and it is ... like ... so millennial ... you know ... and I picked that.

The county intersection is on the far left, and there are three lime green areas where the boundary has moved north.



The census bureau made an effort before the 2010 census to make sure their coordinates were correct. Some of their data was derived from paper maps, which might have had errors, or been careless in choosing datums, etc.

With satellite imagery, they can get more accurate positions for street locations. And they used hand held GPS devices to validate locations of street addresses. If you are near 123 Elm, you can get a location, which can be matched with the block definition (you want to make sure that the address is counted in the correct block). I don't know how they would have caught the Buncombe county line.  Topo sheets can be wrong. Maybe the county GIS identified the mismatch and forwarded it on.

[jimrtex]

You need to try to fashion precise definitions for your categories. It all seems too subjective to me. I always try to formulate such precision in categories in my mind as I think this through, and what seems right, and what will be the impact on maps. I am less concerned with COI issues per se, except as a by product for rules that can be applied objectively, without having to otherwise make such judgments.

Our goal is to create districts comprised of whole contiguous counties, or at least multi-district regions comprised of whole contiguous counties, which can then be subdivided in ways that divide few counties.

A secondary goal is to respect communities of interest. If there is not a community of interest, then it is difficult for a legislator to represent the district, or be representative of the district.

One community of interest are metropolitan areas, which are recognized as Urban County Clusters, the densely settled core of the metropolitan area.

Neighboring counties may also form a community of interest, particularly if residents can easily travel to the other counties for work, buying goods and services, etc. Compact districts are also easier to represent, since there will be fewer media markets, and perhaps need for fewer district offices, or at least district offices that are relatively close to constituents.

Ordinarily, contiguous counties can be considered to be neighbors. But there are exceptions:

(a) Counties which are only superficially contiguous. This includes point contiguity or near-point contiguity. Generally, in these cases, there are intermediate candidates which are neighbors to one or both of the superficially contiguous counties.

(b) Counties where it is difficult to travel between. The community of interest is based on the residents of the counties being able to travel back and forth, where they may interact with one another. They are likely to have different media, there would be less need for government cooperation, and a district office in one, would not be accessible from the other.

Definitions:

Adjacency Index: Length of the border gap, divided by the square root of the area of the smaller county. The border gap is the distance between the end points of their common border. In making the calculation, external waters of the United States (oceans and Great Lakes) should be excluded. If the two counties share only a water boundary, the adjacency index is 1.00.

Quickest Route: The quickest surface route between two locations. The route must be a year-round route accessible by automobile, or passenger ferry.

Circuitous Route: A route is circuitous if the travel time in minutes is more than twice the great-circle distance in miles between the end points.

Direct Route: A route is direct if:
    (1) it it is not circuitous; and
    (2)(a) it is entirely within the two terminating counties; or
        (b)(i) less than 1/3 route is within the intervening county (-ies); and
             (ii) the initial portion of the route would not form part of the quickest route to more than   
                  1/3 of the residents of the intervening county. This calculation should be done in both
                  directions.

Example: A route between County A and County B goes through County C, but less than 1/3 of the route is in County C. Now consider the portion of the route from County A to the A-C county line. If this is part of the quickest route to more than 1/3 of the residents of County C it is not a direct route between A and B. It is an indirect route between A and C, then on to B.

For example, going between Rowan and Stanly on US 52, less than 1/3 of the route is in Cabarrus County. In addition most trips between Rowan and Cabarrus would not use that portion of US 52. They would go south from Rowan or west from Stanly.

A ferry route between two terminous counties is considered to be direct, regardless whether it crosses waters of other counties.

Total Trips The collective set of quickest routes from each resident in County A to each resident in County B. For calculation purposes, a sample may be calculated. Alternatively, all residents of a block group may be be considered to be colocated at the the nearest street to the center of population.

Unconnected and Connected It is easier to define what it means to be unconnected, than connected. Most contiguous counties are connected. Connected counties are contiguous counties that are not unconnected.

Two counties are unconnected if:
    (1) they are not contiguous; or
    (2) the adjacency index is less than 0.05; or
    (3)(a) the adjacency index is between 0.05 and 0.25; and
         (b) the share of total trips that are direct routes is less than 1.075 - 3.5 * adjacency_index
    (4) (a) the adjacency index is greater than 0.25; and
         (b) the share of total trips that are direct routes is less than 0.20 (20%).

The governing bodies of two contiguous counties with an adjacency index greater than 0.05, but unconnected counties may resolve that the two counties are connected, if they believe there is a community of interest between the counties based on factors such as commuting patterns, school districts or urban areas spanning the county boundary; etc. The governing bodies may not resolve that connected counties are unconnected.

I am aware of a Dunn County, Wisconsin, and a town of Dunn in NC. That's it.

Dunn, NC has its own micropolitan statistical area, because it is the largest town in Harnett, County, and about three times larger than the county seat of Lillington. Lillington is the fourth largest town in the county. This is a reason not to concentrate on the county seat.

[jimrtex]

On the Buncombe county GIS web site, the ERSI map layer shows the 4-corner point, but the county layer doesn't match and has an offset. I'm not sure I would know which to use without meets and bounds. In any case we agree that there would be no connection at that point.

So does the the Henderson County GIS web site.

North Carolina statutes say:

G.S. 153A-10 North Carolina has 100 counties and lists their names;
G.S. 153A-17 Specifies that the county boundaries will remain the same unless they are changed
G.S. 153A-18 Provides a procedure for resolving uncertain or disputed county boundaries.

Since the state border between North Carolina and South Carolina was still being resolved recently, I suspect that there are lots of little discrepancies in boundaries.

[jimrtex]

I go with Muon2 on whether to use the county seat or largest city. Some suburban counties have a county seat in the middle of the county, but the largest city is on its edge where the growth from the core UCC county has impinged. Not good. Take a look at Dallas County, Iowa. Adel is the county seat. Waukee is the largest city (the fastest growing city in Iowa actually). In addition, from census to census, as populations change, the county node might bounce around, forcing changes in the map that otherwise might not have to be made. Stability is a good thing, not a bad thing.

One can readily use a distributed location for population. There is no need for a single point. Remember districts represent people, not hogs or corn. Over time, relationships among counties can change. Transportation routes change. For example, Dallas is now part of the Des Moines UCC.

If stability were such a good thing, district lines would never change.

No mills in Waukee?

Oh by the way, the courthouse is not the seat of county government. The building where the board of supervisors meet, is the seat of county government. In Hudson, that building is on State Street, while the courthouse is on the other side of Warren Street on Union Street about six blocks away. So think about that one. Is the Supreme Court building the seat of the US government? Talk about an imperial judiciary!

In many New England counties, the county primarily served a judicial role, and there was no executive or legislative authority above the town. That is why it was so easy to dissolve counties in Connecticut and most of Massachusetts.

In Texas, would the seat of government be where the Commissioner's Court headed by the County Judge meets?

[jimrtex]

I thought my definitions were clear that I do not use the courthouse, I do measure the node from the building where the government meets.

As I was working on CO, I found an interesting puzzler. Consider Pitkin county, home to Aspen. I'm familiar with it and the roads I'll describe having vacationed and had conferences there a number of summers.

To the east is Lake county connected by CO-82 over Independence Pass. At over 12,000 ft it's the highest paved pass of the Continental Divide in the US, but it's closed in the winter.

To the north is Eagle county. CO-82 crosses the SW corner of Eagle from Pitkin before going into Garfield. From that little segment one can take local roads including unpaved stretches through high (and stunning IMO) alpine meadows of the Sawatch range to connect with I-70 and the county seat of Eagle. So they are locally connected, but only by unpaved roads.

To the northeast is Garfield county and Glenwood Springs is the county seat one would most likely reach in the winter from Aspen along CO-82. As I mentioned that cuts about 3 miles through Eagle county between Pitkin and Garfield. You can get from Pitkin to Garfield on back roads that skirt the north side of the Elk mountains, but that includes unpaved roads, too.

To the west is Mesa county, but the short border is made up of the high peaks of the Elk mountains and there are no connecting roads.

To the south is Gunnison county. CO-133 goes from Pitkin to Gunnison, but there are no connections (paved or unpaved) from the rest of Pitkin to CO-133 without going into Garfield (or Eagle and Garfield if paved). Even so, CO-133 doesn't connect to the Gunnison county seat except over unpaved local roads. To the SE there are local crossings from Pitkin to Gunnison, but they are unpaved and closed in winter.

Pitkin may be such an exceptional case that it isn't worth adjusting rules for, but I thought it interesting enough to include in the thread.

Going from Aspen to Carbondale, Glenwood, or Rifle, the passing through Eagle County is superficial and relatively unpopulated. Pitkin and Garfield are connected.

Going from Aspen to anywhere on I-70 from Gypsum through Vail will also take you through Glenwood. A large share of the route and population are in Garfield. Pitkin and Eagle are not connected, unless the boards of commissioners pass a resolution to that effect. They have a COI based on the ski industry.

[jimrtex]

So, Torie it sounds like you would favor a combination of the blue, green and gold links on my map below. That would exclude the local connections (yellow) where there is no nearby highway to establish a link. You thought that jimrtex's definition of a direct route was unworkable. Do you have a definition of how much corner one can cut and still count?

Torie never commented on my more technical alternative. I assume he is still exploring it.

[jimrtex]

So, Torie it sounds like you would favor a combination of the blue, green and gold links on my map below. That would exclude the local connections (yellow) where there is no nearby highway to establish a link. You thought that jimrtex's definition of a direct route was unworkable. Do you have a definition of how much corner one can cut and still count?

Torie never commented on my more technical alternative. I assume he is still exploring it.

This is your "more technical" set of parameters? Oh my. I have a headache. Off the top however, suppose the length of the border between two counties is de minimus, but has a direct local paved highway link, or a qualifying adjacent state highway nick cut? I am not sure that I like that one. Perhaps the best place to start, is what policy issues do you think are not adequately addressed by the system Muon2 and myself are now trying to fashion as described so far above, as a possible set of parameters to make this all reasonably work?

If the adjacency index is less than 0.05, it is a near corner connection.

Think of two square counties, 5 survey townships square (30 miles on each side).

If B is north of A, then they are clearly adjacent.

But let's start sliding B to the west. until there is only a two township overlap (12 miles). That is 0.40.

Slide it another 6 miles, and there is only a 6 mile overlap, and the adjacency index is 0.20.

Now keep sliding until there is only a 2 mile overlap, and the adjacency index of 0.067.

Only until the overlap is 1.5 miles is the adjacency index 0.05.

If you are having trouble visualizing this, let me know, and I can show examples.

Or consider Madison County, Iowa. There are 7 adjacent counties. Which, if any would you have problems with as far as drawing districts between?

[jimrtex]

You be the judge. Here is a link to all the maps of all the urban areas in the US. The natural thing to do is check the maps of NC munis in the vicinity of all the yellow and gold lines on my map. They are locally connected counties, so one just has to see if a route through other counties avoids any of those areas.

There does not seem to be any way to pull up a list organized by state, much less a way to pull up a state map of such areas.

Ah, here is a set of NC maps. My very tentative view, just eye balling how much real estate is sucked up by these urban areas, is that your proposed metric is OK, with one caveat. The Charlotte area lines do not cut off the state highway link between Staley and Carbarrus, but it came close. So maybe an exception to the general rule, is that it is OK to cut through an urban area, if no more than one or two or three precincts are involved, at least within UCC's, where this policy issue would typically arise, and potentially frustrate the public policy purpose of the metric. The whole area is an urban area, but the cut is otherwise minimal and incidental to the main purpose of the highway, which is to directly connect either county seats or urban area populations between the two counties without either 1) a state highway connection, or 2) a local paved highway connection that is a direct link between urban areas in the two counties. By direct, I mean that the road design is to create such a connection with as direct a route as reasonably possible, as opposed to just wandering around servicing local rural homes.

I am less concerned about this 1-3 precinct exception for non UCC counties, because the cases where an urban area would effect a blockade for a nick cut that is for a state highway built to service connections between the urban areas of the other two counties are presumably going to be fairly rare.

I assume that the cut link needs to be a state highway, not a mere direct paved local highway. That is probably OK, but that needs to be clarified.

And then, at the other extreme from the UCC urban area exception issue,  we have the Clay County "nick cut" link.  State highway 64 as it traverses through Clay County from Cherokee to Macon just misses the county seat of Clay, Hayesville, and there are no other urban areas in Clay County. Yet the "nick" traverses the length of the whole county. That is a no go. That is not a nick. I am not sure what rule to fashion to deal with that. That the length of the nick highway cannot be more than a third of the length of the county measured from the two points thereof that give the longest length? Or something like that? And by measuring the length of the highway for this ratio, the rule would tend to penalize twisty state highway nick cuts that twist and turn over rugged terrain, which is a good thing I would think. An absolute highway distance parameter not based on a ratio might not work too well, given counties vary a lot in size.

You will want to look at the urban area map for Concord, NC.

I though that Muon2's concern was with placing Rowan in the same district with Stanly.

[jimrtex]

Madison County, Iowa? Never heard of it. Why did you pick that county?

Madison County does not present any interesting issues. All the adjacent counties are linked by state highways, except Clarke County, which does not have a qualifying local paved road qualifying link. It is not direct. So Clarke cannot be in the same CD merely connected by the Madison-Clarke border.

For simplicity, we will consider Winterset and Osceola to be equivalent to their respective counties. This is particularly true for Osceola/Clarke. Madison has had an uptick in population that must include some folks living closer to Des Moines.

I'm not sure whether 1/3 of the quickest route between Winterset and Osceola is in Warren County.I'll assert that is not. The quickest route from Winterset to Indianola, and from Osceola to Indianola don't use the same route.

In any event, the county commissioners might decide that the fact that I-35 is a mile east of the county line doesn't have much significance.

What about Madison-Polk or Madison-Adair?

I understand what you wrote quite well (for once! ), and I still don't find much favor with that constraint. If there is a direct local paved road local link (very unlikely to be one in Iowa for a very narrow border), that is good enough for me. Ditto with a qualifying state highway nick with also a pavement connection (also very unlikely in Iowa).

Below are four adjacent counties. The only urban areas are the county seats, which are represented by the little squares. The thick black line below is a state highway. The thin lines are paved local highways. There is no direct paved road link between counties A and B, but there are paved highway connections. The state highway is a qualifying county nick cut connection. No matter how short the distance of the adjacency line between counties A and B, I consider them connected assuming a local paved road squeezes through whatever real estate is adjacent between the two counties.

I don't see the point of requiring the local paved road.

The adjacency index is 0.09, and the residents traveling from Alphaville to Betaburg are using the highway. They don't use the local paved roads, and if they even travel to Seton or Deep River, they will go some other way.

[jimrtex]

The one third test is about the length of the road that is in the foreign county relative to the length of the foreign county. Adair and Madison have a state highway link. Even if it were not a state highway, while the road has a couple of jogs, it is direct enough really. The road is designed to link the two county seats, and just had to avoid bisecting a couple of farms, where the land owners back when had pull, or something. No problem. (Actually it is because diagonal roads are unpopular in Iowa because everything is cut up into square sections, so if a road has to move a bit north or south to get to the next county seat, it needs to have these jogs to avoiding slicing the sections.) Polk and Madison have but a point connection, so that obviously does not qualify even if one did not have a pavement link requirement. Muon2 and I want to be able to drive on a paved road between the two counties, without having to go through a third county. Muon2 paraphrased that dude who said that metric was important to him. That I guess is quite popular out there on the Fruited Plain.

Actually, I meant to ask about Guthrie and Madison, rather than Adair and Madison.

The dude likely had not fully thought through the issue, or was not thinking about using whole counties.

The population increase in Madison is almost entirely in Winterset is my impression. Next to no housing tracts have been built in the NE corner of Madison yet. But Winterset has some new subdivisions. Not to the south however, because that is where the Torie farm is, and Torie isn't selling yet.

Between 2000 and 2010, the 7 townships on the north and east county lines (near I-80 and I-35) increased by 893. Winterset and the 4 surrounding townships 727, and the remaining 5 townships on the south and west sides lost 70.

If you want to live in the country, you can have a job in Des Moines and commute to a small tract of land that would be too small to support full time farming.

[jimrtex]

The one third test is about the length of the road that is in the foreign county relative to the length of the foreign county.

Under my criteria, the 1/3 is relative to the length of the route.

Muon2 and I want to be able to drive on a paved road between the two counties, without having to go through a third county. Muon2 paraphrased that dude who said that metric was important to him. That I guess is quite popular out there on the Fruited Plain.