US House Redistricting: Tennessee

[muon2]

Here's my whole county version of the map. All districts are within 0.5% of the ideal, and the maximum deviation is 2506. CD 9 is entirely within Shelby and is 56.0% BVAP. At least one county must be split around Nashville to stay within the population range, and Davidson was split since it is the largest.

[muon2]

Here's my whole county version of the map. All districts are within 0.5% of the ideal, and the maximum deviation is 2506. CD 9 is entirely within Shelby and is 56.0% BVAP. At least one county must be split around Nashville to stay within the population range, and Davidson was split since it is the largest.

Hey, Mike, if this were in the Midwest, your TN-05 would have a GOP leaning PVI!  But TN isn't in the Midwest, so it doesn't. Darn!  

In any case it's pretty easy to change the grouping of CD5 and 6 to create two districts each with about 51% McCain and 53-53% Rep in the DRA. For example, The six counties can be split as below without losing much compactness. CD 5 is 51.6% R, 54.9% McCain; CD 6 is 51.0% R, 53.8% McCain.

[muon2]

In every election there is a winner and a loser. The people whom vote for the loser aren't as happy with the result as those that vote for the winner. No matter how you divvy up the lines, elections will result in large numbers of voters voting for the loser.

For you to specify one group of losers are being particular aggrieved by backing the losing candidate is just another example of using self-serving standards.

They are not aggrieved by backing the losing candidate. They are aggrieved because the mapmakers decided that those people in particular should be the ones who back the losing candidate.

You must be logged in to read this quote.

Again, the issue is not that people back losing candidates, it's that the mapmakers effectively determine who the losing candidates are, and by extension, deny representation to the voters who back those candidates.

This was the core of an argument made by the IL League of Women Voters in their attack on the legislative and congressional maps drawn by the Dems this year. It's an argument that has lost in the past at SCOTUS, but not without sympathetic words from the justices. This year they took the novel step of attaching that argument to the recent Citizens United and Freedom Club PAC rulings. Unfortunately, the federal district court panel had their case consolidated with the GOP state case and then their claims were dismissed. There may yet be further legal action along this line.

[muon2]

In every election there is a winner and a loser. The people whom vote for the loser aren't as happy with the result as those that vote for the winner. No matter how you divvy up the lines, elections will result in large numbers of voters voting for the loser.

For you to specify one group of losers are being particular aggrieved by backing the losing candidate is just another example of using self-serving standards.

They are not aggrieved by backing the losing candidate. They are aggrieved because the mapmakers decided that those people in particular should be the ones who back the losing candidate.

You must be logged in to read this quote.

Again, the issue is not that people back losing candidates, it's that the mapmakers effectively determine who the losing candidates are, and by extension, deny representation to the voters who back those candidates.

This was the core of an argument made by the IL League of Women Voters in their attack on the legislative and congressional maps drawn by the Dems this year. It's an argument that has lost in the past at SCOTUS, but not without sympathetic words from the justices. This year they took the novel step of attaching that argument to the recent Citizens United and Freedom Club PAC rulings. Unfortunately, the federal district court panel had their case consolidated with the GOP state case and then their claims were dismissed. There may yet be further legal action along this line.

The argument might fly for classes of people, and, seems to have been legislated for minorities in the VRA. Arguments that protect zip codes, not classes of people, are going to go nowhere.

And, as I noted before, the notion of members of political party being deemed a protected class is outrageous and dangerous. I'd rather lose in Illinois that throw in with the notion that the government, and the courts, should decide the partisan composition of legislatures rather than the voters.

I would largely agree if it were the case that the voters were choosing the partisan composition of the IL legislature. In this case the government in power chose the partisan composition for the next decade by their map. There were no checks by the people on that use of government power. Even in OH and MI there are some checks imposed on the government's power to lock in a specific partisan result. In IL there are none.

[muon2]

Just for kicks, I did a Pub gerrymander light for TN, which creates a nice competitive CD for the TN-05, which should make the AZ "non-partisan" redistricting commission proud. Its underlying leitmotif is that it keeps the black community of Nashville together (a few inner white liberal precincts are excised in partial exchange), and does but one split of Davidson County. It also makes sense in other ways I think, with the Nashville metro area divided between 2 CD's.

Your western TN and Nashville metro area looks similar to my whole county version from a year ago.

Here's my whole county version of the map. All districts are within 0.5% of the ideal, and the maximum deviation is 2506. CD 9 is entirely within Shelby and is 56.0% BVAP. At least one county must be split around Nashville to stay within the population range, and Davidson was split since it is the largest.

Since we've been discussing erosity as part of a series of factors, I decided to examine my plan from that perspective. This is what the link map of TN looks like.



The plan has seven regions (2 divided counties) and at the regional level splits 63 links for its erosity score. That increases to 68 for the internal splits of Shelby and Davidson, or 69 if I use Torie's split of 5 and 6 instead. The maximum range for the plan is 0.8% at the regional level and 0.59% at the district level. Microchops of less than 0.5% of a district can be used to get exact equality if desired.

I've done a little rearranging to see if either the erosity or range could be improved while keeping the plan at seven regions. This version has an erosity of 60 at the regional level and 63 including county splits. The range goes to 0.7% by region and 0.52% at the district level. Since it is better on these measures and there are the same number of county splits this would be Pareto preferred to my plan of a year ago.



I wasn't drawing it to give a particular partisan result, just following the river and major roads to the extent possible across Nashville. My CD 5 is 54.1% McCain and CD 6 is 47.0% McCain. Perhaps Torie would like to see if that tweaks as well to give his desired gerrymander.

[muon2]

Oh Muon2 your map above might work as a Pub gerry-lite, but I must say my tweaked map below appears to me to be superior to your algorithmic map. TN-5 is 50.0% McCain, 48.8% Obama, and TN-06 is 57.3% McCain, 41.6% Obama.



Your cut of Davidson appears to slice the black community by the way. I assiduously avoided that in my map (the NW black node is entirely in TN-05, albeit a handful of substantial black minority precincts in the geographically separated by white precincts SE node are in TN-06). And notice how it keeps downtown Nashville whole, and uses that lake as a barrier on the east side. Natural barriers are there to be used where possible - not ignored.




What do you think?  I am still a little confused btw what the real purpose of the regional overlay is in your formula, above and beyond county cut containment. Is to to take cognizance of metro areas, and keeping them whole, ultimately, or is there more? Is it a way to contain erosity, or limit erosity to where it ties together "regional" communities of interest?  In any event, everyone "knows" the regions of Tennessee, including most of its citizens, and my map took that into account. I like there being one middle TN rural CD, TN-04 (with rural TN-07 being 75% Middle TN, and 25% West TN by necessity to equalize population), and two CD's that cover the Nashville metro region, and no more, and fit like a glove over that metro area. That is how it should be.  I also tried to limit just visual erosity, and move towards rectangular squarish shapes. And I looked at road ties in East TN, and moved stuff around a bit accordingly, including moving Sevier County into the Knoxville CD, where it properly belongs really, rather than TN-01.

With some tweaks your plan is an appropriate one to compare to mine. I'll start by removing microchops which can always be added at the end if exact equality is required. Then, by the regional approach you have essentially placed the three eastern CDs into one region. To make that clear I have removed the chop in Lincoln county by putting all of Marshall into CD 4. I also removed the chops in Anderson and Claiborne by moving Rhea into CD 4. In the Knoxville region I then made the two required chops in a way to minimize internal erosity.



Overall the plan has a range of 0.8% at the region level and 0.7% at the district level. There is an erosity of 42 at the region level and 54 at the district level. Compared to my plan yours increases the range and county splits but improves erosity. That's an appropriate Pareto shift so both could be considered. A plan is excluded only if one can make another plan that improves on any one variable (splits, range, erosity) while make the others no worse.

[muon2]

Muon2, how did you define regions for TN (population driven, or historically driven, or both?), and the road splits only count when a CD splits a region?  I still don't quite get the mechanical aspect of regions in your methodology.

A region is a connected set of counties that has a population nearly equal to a whole number of districts. The use of roads determines connectedness, and the severed connections also serve to measure erosity. When a county is split within a region then each piece is treated as if it were a separate county for connections and erosity.

The main idea is to recognize that split counties (or cities) and erosity both point at potential gerrymandering and generally point away from maintaining communities of interest. My goal is to create a mechanism whereby those two factors can be balanced against each other and balanced against population equality.

[muon2]

Muon2, how did you define regions for TN (population driven, or historically driven, or both?), and the road splits only count when a CD splits a region?  I still don't quite get the mechanical aspect of regions in your methodology.

A region is a connected set of counties that has a population nearly equal to a whole number of districts. The use of roads determines connectedness, and the severed connections also serve to measure erosity. When a county is split within a region then each piece is treated as if it were a separate county for connections and erosity.

The main idea is to recognize that split counties (or cities) and erosity both point at potential gerrymandering and generally point away from maintaining communities of interest. My goal is to create a mechanism whereby those two factors can be balanced against each other and balanced against population equality.

That statement suggests regions serve no purpose whatsoever, unless road splits that hew to regional lines don't count, or regional lines must be followed except to round up or down to whole CD's in the case of the one CD that will cross regional lines. It seems more a matter of convenience in calculating where CD's will end up going. And what is to prevent erose regions, if regions are strictly driven by population?

It is certainly true that I could set up the same measures directly to the district level. In that case the regions would serve to guide the mapper towards possible county groupings that would likely do well in the final scoring. That's actually the way I used them when I drew my OH maps that did well in competition, since there was no score or knowledge of my intermediate steps.

If one does require a Pareto optimization of regions as well as districts it provides a constraint on the mapper that should tend to force better connected groups of counties together. That's a plus for improving maps where connectedness is presumed to relate to communities of interest.

It is also the case that the computerization of redistricting remains hard, and for the number precincts (far worse for census blocks) the problem is sufficiently hard that the best maps would be difficult to draw in reasonable times. By splitting the problem into two tiers - one for regions and one for districts, the computational problem becomes exponentially easier. This also implies that human mappers would have to focus on a reduced set of options providing a key constraint against hidden gerrymandering.

You must be logged in to read this quote.

For the four states I've done (WI, WV, WA and TN) I used a paper atlas, Mapquest, and Paint to create the links. It takes a couple of hours. It's easy to visually count the cuts once I g=have a map, but it would also be straightforward to put the links in a table and have a simple program measure the plan. One could make a blank county map and have users color code them just like election prediction calculators do for region construction.

You must be logged in to read this quote.

I presume that real splits would use data at a finer level than precincts. That's certainly needed to avoid chopping cities when precincts overlap the boundaries. States generally redraw precincts after a remap causes splits anyway.


But let me assume that for municipal split purposes you really want a district boundary to follow the image above. If 70S continues along the boundary to the west and eventually to the population center for the blue piece of the county then there is a link between the blue part of the west county and the east county. That link would count for erosity.

In your Nashville plan, there is no link from north Davidson (CD 5) to Rutherford county (CD 6) north of the lake, leaving only south Davidson linked to Rutherford, but that's an internal link not counting for erosity. However, as you drew it there are links from both north and south Davisdson to Cheatham county so that increases the erosity by one more than would otherwise exist on the border of those counties.


In my tweak to your plan I shift the boundary south of I-40 so there is no link from south Davidson to Cheatham, but south Davidson remains well connected internally along 100.

[muon2]

Yes, there is no road split between Cheatham and South Davidson, but one more road split between North Davidson and Cheatham, so why one less road split with your revision of my map?
 I am confused again. In all events, as an factual matter, there really is no greater erosity or lack of connectedness, but until I understand your program clearly, there is no point going there.

My general rule is that when a county is chopped, each chop part now acts as a new county for the purpose of connections and thence erosity. Connections between counties depend on a path of numbered federal and state highways between population nodes. The population node is determined by the largest urbanized area (census maps) or the county seat if there is no urbanized area, or the largest population incorporated or census designated place if none of the other choices are available. Within an urban area the largest population within the county is used for the node. This can be either incorporated or unincorporated, or the downtown within a large city (typically including city hall).

It's easier to start with examples from eastern TN. This is my split of Grainger. There are no urbanized areas in Grainger so county sear Rutledge is the population node. It stays in the northern chop and Blaine is the largest place in the southern chop, so it is the population node for determining connections.

Grainger (Rutledge) is connected by state or federal highways to to Claiborne, Hancock, Hawkins, Hamblen, Jefferson, Knox, and Union. The split of the southern part around Blaine is only connected to Knox, and the remaining northern part is connected to the remaining counties, but not Knox. This is an ideal chop that has no increase in erosity.



This is my split of Blount. Much of the population of the county is within the Knoxville urban area which doesn't show on the DRA map. Maryville is the largest place and within that urban area and is the population node for the county which is connected to Monroe, Loudon, Knox, and Sevier. The northern part including Maryville is only connected to Knox and Sevier.

The southern part has the largest population in the urban area in unincorporated Binfield, and the largest place is the much smaller Friendsville. Binfield has the larger population in the urban area and is used as the node connecting to Monroe and Loudon. Friendsville only links to Loudon, but a connection that exists at the county level cannot vanish by the means of a district line, so the Monroe connection would have to remain there as well. This chop also separates the county connection in a way that does not increase their number.



Davidson county is problematic to define since Nashville includes all of the county not in the few small communities. The northern part has the downtown so that defines its node. The southern part has more population in Nashville than in the smaller separate cities and the largest part is along I-24 so that defines the node for connections.

Davidison connects to Cheatham at the county level, but doesn't need to have connections to both the northern and southern chops of Davidson depending on where the line is drawn. By following the river at the Cheatham line you provide a connection to the south Davidson node along TN-1, TN-100 and TN-254. By moving the line south as I do I remove the link from Cheatham to south Davidson. If I had controls I'd make my cut more cleanly than the course VTDs in DRA.

Yes, it may seem arbitrary that our two plans seem reasonably equal by eye in Nashville, yet differ by this measure. The difference is that without my constraint on erosity, a mapper is free to place the line at Cheatham anywhere they want and that opens the door just a bit to gerrymandering. The idea is that a reasonable set of constraints can limit gerrymandering yet provide flexibility.