House breakdown if...

[muon2]

Definitely a few more Democratic districts. The problem is that if you draw districts normally, there's still a lot of wasted votes in overkill Democratic areas like the west Bronx.

I've highlighted the critical part of your post. It's absolutely correct, and in general will weigh against the first part of your quote. For instance, I went through a straight population exercise for IL a few months ago in this thread. If anything it reduces the number of secure Democrat districts by eliminating the gerrymander along the Mississippi (IL-17) and compacting the Chicago districts.

It does increase the competitiveness of many of the districts. Results like Bean's win in IL-8 would be more common. But Republicans would have an equal shot at some lean Democrat districts as well.

[muon2]

Definitely a few more Democratic districts. The problem is that if you draw districts normally, there's still a lot of wasted votes in overkill Democratic areas like the west Bronx.

I've highlighted the critical part of your post. It's absolutely correct, and in general will weigh against the first part of your quote. For instance, I went through a straight population exercise for IL a few months ago in this thread. If anything it reduces the number of secure Democrat districts by eliminating the gerrymander along the Mississippi (IL-17) and compacting the Chicago districts.

It does increase the competitiveness of many of the districts. Results like Bean's win in IL-8 would be more common. But Republicans would have an equal shot at some lean Democrat districts as well.

Democrats would still be better off, as more states have pro-Republican gerrymandering (like Michigan, Pennsylvania, and Texas).

It's an interesting hypothesis that's worth testing. I'm working on a set of rules that a presumably independent commission would follow to create a map. I would start with jimrtex's rules, then modify them accordingly.

I find there are two types of states for which one can set up rules. One set, like TX, have no county subdivisions recognized federally other than census tracts. Other states in the north and east have towns or townships as county subdivisions. It's easier to create and apply nonpartisan rules to those states with county subdivisions, so let's start there.

In the jimrtex rules, there are three primary classes of division to avoid. there are other rules, but these seem to dominate the map creation when I've look at a few example states.

1) Splits of the units (county or town) which are less than the population of a district.

2) Keeping at least one whole district in units larger than population of a district.

3) Creating multiple partial districts in any split unit.

I'd like to invite thoughts on what priority order to apply these rules.

[muon2]

Ok, I have my first pass at some states using purely geographic rules to create CDs. I defined rules for states with well defined county subdivisions.

Definitions
D-1: A town is defined here to include town, township, village, and city subdivisions as recognized by the Census Bureau as county subdivisions.

D-2: Discontinuous parts of towns shall be treated as separate towns for each contiguous section.

D-3: Areas may refer to towns or counties and their boundaries are as defined by the Census.

D-4: A partial district is a portion of an area less than the whole, containing all the population of that area in a particular district, and contains less popluation than needed for a whole district.

Fixed Rules
F-1: The size of each district shall be within 0.5% of the average size for the state.

F-2: Each district shall be contiguous.
F-2-1: Contiguous shall not include contact at one corner only.
F-2-2: Contiguous shall not include a district with two areas on the same land mass that are contiguous only by means of crossing a body of water, except when those two areas share a common boundary in that body of water.

F-3: Each plan for a state will be judged based on a set of priority rules. The highest priority (lowest numbered) rule takes precedence in determining the composition of the districts for a state.

Priority Rules
P-1: The fewest towns are split that have less population than the state average.

P-2: The fewest towns split that have a population greater than the state average such that there is no district wholly within the town.

P-3: The fewest total partial town districts.

P-4: The fewest counties are split that have less population than the state average.

P-5: The fewest counties split that have a population greater than the state average such that there is no district wholly within the county.

P-6: The fewest total partial county districts.

P-7: The fewest partial county districts in any one county.

[muon2]

I found that the rules force one to start by creating districts that can include the largest counties less than the population of one district. The last districts created are those in counties with more than enough population for one district, because they will have the partial districts needed to bring other districts within the population guidelines.

For my first application of the above rules, I have redivided MN, WI, and MI.


Some comments on the above map.

MN has zero for each of the three town rules. Rule 4 has one split in Ramsey, due to the fact that keeping it intact requires a split of Minneapolis which is worse for rule 1. Rule 5 has zero, rule 6 has four partials, and no more than two partials under rule 7.

WI also has zero for each of the three town rules. The City of Milwaukee isolates the north shore suburbs in its county from the rest of the towns. Those suburbs can't be grouped with the district north of Milwaukee to make its population work out, so I had to split Washington County (I separated West Bend as one of many options).  The result is one split under rule 4, five total spilts under rule 6, and three splits maximum for rule 7.

MI has three counties bigger than the district minimum, and one town larger. This allows rules 1, 2, 4, and 5 to be at zero. Detroit requires that rule 3 have one town split. The shape of Detroit isolates groups of towns in Wayne County on the East, and the border towns in Macomb have too much population, so Oakland is stuck with four partial districts.  Overall MI has six partials.

[muon2]

Now let me address the question of competiveness and partisan leanings in those three states compared to this "non-partisan" map. To do this I will use the percentage of the 2004 presidential vote given to the major party candidates. This clearly won't account for incumbency and other local factors, but can be used to judge the district.

A competitive district has less than 52% to the leading party.
A leaning district has 52% to 55% to the leading party.
A strong district has more than 55% to the leading party.

Currently in MN there are 5 Republican (2 strong, 1 leaning, 2 competitive) and 3 Democrat (2 strong, 1 leaning) districts. With the remap, there would be 4 Republican (3 strong, 1 competitive) and  4 Democrat (2 strong, 1 leaning, 1 competitive). The competitive district that switches is in suburban Hennepin and goes from 51.45% Republican  to 50.27% Democrat.

Currently in WI there are 4 Republican (2 strong, 2 leaning) and 4 Democrat (2 strong, 2 competitive) districts. The remap keeps 4 Republican (3 strong, 1 competitive) and 4 Democrat (2 strong, 2 competitive). Not much change happens here, perhaps one additional competitive district in the SE at the expense of a stronger one in the NE.

Currently in MI there are 10 Republican (3 strong, 5 leaning, 2 competitive) and 5 Democrat (all 5 strong) in a state that was questioned as being overly partisan in its gerrymander. The remap gives 8 Republican (4 strong, 3 leaning, 1 competitive) and 7 Democrat (4 strong, 2 leaning, 1 competitive). The density of Democrats in Detroit still leaves a majority of districts on the Republican side. However, the strong district 12 in southern Oakland and Macomb and the strong district 15 in Ann Arbor transfer enough votes to flip suburban Oakland and Macomb districts to strong D and competitive D respectively, leaving a leaning D district in suburban Wayne.

[muon2]

My next set of states are IL, IN, and KY (yes I'm recycling the maps used by jimrtex in his Senate project.
 


Here are my comments on the above map.

IN used exactly the rules I listed earlier. Priority rules 1 through 5 all come up zero. Only Marion is split and exactly two whole townships each add to other districts, thus rule 6 and 7 have two partials.

KY has no county subdivisions, so rules 1 through 3 don't apply. Rules 4 and 5 have zero splits, and only Jefferson is split with one partial district for rules 6 and 7.

IL is tough around Chicago. Chicago is a county subdivision, but large enough for more than 4 CDs. Fortunately there are well established Census subdivisions within Chicago, so I treated these "community areas" as towns within their county.  With this new definition, I needed no split towns for rules 1 through 3.  I did need to split three "collar" counties (McHenry, Kane, and Will each split into two partials) around Cook for rule 4 and gave DuPage no fully contained district for rule 5 (it has 4 partials). Cook is also left with 4 partials, so rule 6 has a total of 14 partials, and rule 7 has a maximum of 4 partials in any one county.

[muon2]

And here is the political analysis of those districts, using the 2004 vote in each county or town. I approximated where the voting data was not a perfect match to the town boundaries.

As expected heavily GOP IN and KY show up that way in these districts as well. IN has seven strong Rep districts, one strong Dem district in NW IN, and a competitive Dem district (51.9%) in Indianapolis. KY has five strong Rep districts, and one competitive Dem district (50.8%) in Louisville.

IL is one of the most heavily gerrymandered states, designed to have 10 Rep and 9 Dem districts. In this rule-driven plan each side has seven strong districts (well actually two Rep districts are at 54.6 and 54.8%, which only round to 55%.) An interesting battle would be in Chicago's new CDs 4 and 7, which are gerrymandered to separate the Black areas from the Hispanics. The new split could easily produce two Black or two Hispanic members in some nasty primary fights.

The other five are very competitive and three would have voted Dem in 2004. They are:

New 8 (50.4% Dem): Now entirely in north and northwest Cook it is centered around Schaumburg, Arlington Heights and Palatine, but goes all the way to the lake.

New 10 (50.9% Rep): All of Lake and a corner of McHenry. The real IL 10 is probably one of the most potentially competitive districts, but Kirk has done well there.

New 11 (50.9% Dem): This moved into the suburbs of south Cook and SE DuPage. Joliet and Cook is balance by the rsst of Will and Downers Grove.

New 12 (50.9% Dem): This reliable Dem seat centered on East St. Louis suddenly turns competitive as it picks up most all the St Louis suburbs and a lot of GOP rural area.

New 16 (50.1% Rep): This area now combines Rockford and Rock Island, and gone is the bizarre shape of real IL-17. Peoria moved from 18 into the new 17, and Springfield is centered in the new 18 without a slice right through its middle.

[muon2]

And here is the political analysis of those districts, using the 2004 vote in each county or town. I approximated where the voting data was not a perfect match to the town boundaries.

As expected heavily GOP IN and KY show up that way in these districts as well. IN has seven strong Rep districts, one strong Dem district in NW IN, and a competitive Dem district (51.9%) in Indianapolis. KY has five strong Rep districts, and one competitive Dem district (50.8%) in Louisville.

In your plan, or currently? Or both?

My plan was based solely on the rules and I checked the political data only after completing the map. What I expected was that since both IN and KY voted about 60% for Bush, and that vote was widely spread across those staes, the mock CDs would show similar strength for the GOP.

The real IN delegation matches the mock CD expectations exactly. The KY delegation matches the number but not the distribution, since Northrup (R) holds the Louisville seat, and Chandler (D) holds the Lexington seat.

[muon2]

The competitive district that switches is in suburban Hennepin and goes from 51.45% Republican to 50.27% Democrat.

*falls off chair with happiness*

And Rice ended up in BRTD's as well. Double bonus.

I'm working on the other northern states that have defined county subdivision.

[muon2]

Waiting...

... not for very much longer.

I should have a new set up this weekend. This has been a very busy week between teaching, official meetings, parades, and fundraisers.

[muon2]

Here's some conversation on the issue of where these maps come from, from the related thread "Look what he's done to US!"

Can I ask how you made those excellent maps?  Where did you find the source maps?

I found a national county map on the census bureau site in PDF format (the large scale projection is what causes the eastern states to tilt upward to the east).  Anyhow, I can zoom in on the region of interest and then I do a screen capture.  I then spend too much time editing with MS-Paint (an artifact of the PDF file is that river meanders are digitized at a resolution higher than I'm displaying on my maps.  So they end up as black splotches unless I edit them.

The advantage of doing it this way is it helps when figuring out the cross-border districts, since everything is at a common scale and relative position.

There are a set up state level base maps on an Indiana State University site which I can't find right now, which are a lot faster for single states.  The census bureau also has PDF files of each state.

Edit: found URL.

State basemaps from ISU

I would STRONGLY recommend you use Adobe Photoshop instead of Paint.  You can do the same job a lot more quickly and easier, and you won't have the same resolution problems.
It's well worth the money, but I guess if you don't want to pay then that's not a problem, because you can find a free copy on Limewire or Kazaa.  Or, PM me and I might be able to hook you up too bad.

Can it edit (color fill) a PDF file?

Yeah, it opens pretty much any kind of image file.

Lewis is correct. At this point I'm just editting other image files. For simplictity I like gifs or something I can pull into Paint. The ideal circumstance is colors with no shading so its easy to recolor as needed. jimrtex's senate maps have worked well so far.

[muon2]

I thought I posted this yesterday, but I must have erred somewhere ...

Here is New England carved into CDs by my geographic rules:

You must be logged in to read this quote.

The good part was that the large number of towns in New England made it easy to avoid any splits at that level, so P-1, P-2, and P-3 are all kept at zero. Even Boston fits within a district. The hard part was that there were relatively few counties, and many with large populations. This forced more split districts, particularly in MA. Also MA has one county (Norfolk) that has two towns disconnected from the rest of the county, so I extended definition D-2 to apply to counties in a similar way.



My comments:

ME convieniently is divisible into two districts entirely along county lines (0.47% deviation for each). So, all priority rules equal zero.

NH requires one county to be split, so I selected Hillsborough which is the largest. The result is P-4 equals one, P-6 and P-7 equal two.

VT has one district and works out perfectly.

MA was quite difficult. There is only one contiguous combination of towns with Boston that falls within the 0.5% limit: Winthop and Milton. This forces extra splits in Suffolk and Norfolk. The population of Barnstable plus Plymouth is just slightly too large for one district, and adding Bristol and the islands is just barely too small for two districts,. This forces splits in Plymouth and another partial district (one town) in Norfolk. It was possible to put a complete district in each of the other large counties (Worcester, Middlesex, and Essex) and split no other small counties. The result is P-4 equals two (Norfolk remainder and Plymouth), P-5 equals one (Suffolk), P-6 equals thirteen, and P-7 equals three.

RI has one county in excess of a district (Providence) and it gets a whole district plus a partial. There were no other splits, so P-6 and P-7 both equal one.

CT has three counties in excess of one district and each got a whole district. With no other splits, P-6 equals four and P-7 equals two (Hartford).

[muon2]

Reminds me of that Mass. map I once drafted...with that Lowell-to-West MA district, the districts based on Springfield and Worcester, and also in the SE part of the state.

I really wasn't happy with the Lowell-Northampton district, particularly since it goes deep into Essex as well. However, I could find no other way to have a complete district in both Worceseter and Middlesex. As I noted in another thread a while ago, better maps in MA would ignore county lines since they aren't significant political divisions compared to towns. But, I kept to my rules and was left with the long northerly district.

[muon2]

the MN map shows why always not trying to split counties is not neccesarily a good thing. You have St. Paul suburbs in with Minneapolis and its suburbs, which would not go over too well, and you have Washington county seperated from the rest of the metro and in with some rural area instead.

However, I would like this map better since more Dem-leaning districts are created, especially us getting Rice back

See, I just knew I could find something to make you happy.

Actually, your point is very accurate. Limiting gerrymanders through any geographic constraints will put together populations that may traditionally not go together. In Iowa this happens every ten years since they have a hard restriction against splitting counties, and now use a computer-based process for their CDs.

[muon2]

Actually, your point is very accurate. Limiting gerrymanders through any geographic constraints will put together populations that may traditionally not go together. In Iowa this happens every ten years since they have a hard restriction against splitting counties, and now use a computer-based process for their CDs.

They also use an odd definition of compactness - basically that the east-west extent should be similar to the north-south extent.  But it means that an "L" shape is just as good as a rectangle, which is shown by Districts 1, 2, 4, and 5.  Going further east in the southern part of 5, counts as increasing the east-west extent measured from the NW corner rather than from the SW corner.

A more traditional definition would consider the ratio of the circumference of the shape to its area -- lower is more compact. I suspect that the more traditional measures were a bit too mathematical for a legislature to be comfortable in passing. I also chose not to use any compactness rule for my maps. I am willing to allow some political flexibility, just within limits to restrict the most egregious gerrymanders.

[muon2]

I've completed Ohio using the usual rules.



My comments:

Only one city, Columbus, is larger than the population of a district. Columbus also has a number of unincorporated pockets inside, and I'm only guessing that I can make a district entirely within the City limits, and provide the necessary links between the pockets and the adjacent districts. With that assumption, rules P-1,2,3 are all zero.

OH has a number of counties with large urban centers, and that forces some county splits.  I split  Montgomery and Stark that were less than one district size, and split  Stark into three pieces to avoid
splitting a third small county.  Rule P-4 is two.

All three large counties had a district wholly within them, and Cuyahoga has two wholly within. Rule P-5 is zero. All three also only have two partial districts. Rule P-6  is 11 partial districts, and rule P-7 is three maximum in any county.

[muon2]

The political result of the above map is quite different than the current set of CDs. Like MI, OH has a strongly tilted map for the GOP. They use a rule that avoids creating more than two partial districts in any county, so it doesnt look so gerrymandered. However, many large cities get split, so the current distribution is 13 R (6 strong, 3 leaning, 4 competitive) and 5 D (all strong). One competitive R district is currently held by a Democrat.

One feature of my rules is that they keep cities intact. In this case cities like Cincinnati and Columbus basically get their own districts turning two competitive R districts into a strong D and a lean D district. Some of the R districts in the Cleveland exurbs also disappear as the D population speads into more districts. The result of this map is 9 R (6 strong, 2 lean, 1 competitive) and 9 D (5 strong, 3 lean, 1 competitive).

[muon2]

Have you classed the big blue SW district as strong, lean or competitive?
I'td have voted strongly for Bush, but I'd guess would have re-elected Strickland again.

Strickland's district may look like it is primarily along the Ohio, but most of the population lives in Mahoning and Columbiana counies by Youngstown. That includes Strickland and his new district (green in the northeast on the map) would be lean D - 54.1%.

The big blue district in the south would be lean R - 54.6%. It has no incumbent in it. That large rural area is carved up by four districts at present.

[muon2]

These maps are amazing, Muon (as are jimrtex's).  Is there any way you would consider making districts using your rules for Georgia?

Thanks. One issue with GA is that there are no political subdivisions of counties equivanent to towns as exist in the North and Northeast. The Census bureau does subdivide counties in GA, but without some investigation, I don't know if they would be as useful as towns are in the 25 states that have them.

[muon2]

One feature of my rules is that they keep cities intact. In this case cities like Cincinnati and Columbus basically get their own districts turning two competitive R districts into a strong D and a lean D district.

Don't you mean a strong R (suburbs) and a lean D (city proper)?

Cincinnati is currently split between CD 1 and CD 2. By moving all of Cincinnati into CD 1 it goes from competitive R (50.5%) to lean D (53.1%). The shift strengthens CD 2 from 64% to 69%, but it is solid R either way.

Columbus is currently split between CD 7 (strong R 57.4%), 12 (competitive R 51.6%), and 15 (competitive R 50.5%). By making CD 15 entirely within Columbus it goes to strong D (62.7%). CD 7 gets a little stronger (59.8%) and CD 12 shifts from competitive to strong R (57.3%). This is an unusual case where competitiveness decreases.


The situation in ne OH is also interesting. Two R districts vanish there as well.

There are 7 CDs that cover ne OH: 6, 10, 11, 13, 14, 16, and 17. My version keeps 10 and 11 entirely within Cuyahoga, and their R-D percentages are essentially unchanged. Four GOP counties near Cleveland form the basis for 2 GOP CDs: Lake and Geauga for CD 14, and Medina and Wayne for CD 16. My map splits those four counties into four different CDs.

CD 6 currently combines Columbiana and part of Mahoning (Youngstown) with a strip along the Ohio River for a competitive Bush district (50.3%). My map puts the river counties with Stark (Canton) shifting it barely to D (50.8%).

CD 17 links the rest of Mahoning with Trumbull, Portage, and the Akron part of Summit to make a very solid D-district (62.8%).  Now I link Mahoning with Portage and Geauga to make a lean-D (54.1%).

CD 14 is one of the current exurban R-districts (52.7%). By dropping suburbs in Summit and all of Geauga, while picking up all of Trumbull this district switches to strong D (55.7%).

CD 13 presently links westerns suburbs of eastern Lorain with much of Summit for a solid D-district (55.4%). My map combines Lorain with medina and three counties to the west to make a district with no partial counties. The switch moves the district to marginal R (50.1%).

CD 16 is the other R-district in ne OH (54.0%). Summit county is currently divided between three districts to make two D (13 and 17) and one R (14). My map completely eliminates the existing CD 16, moving Wayne into CD 4, and creates a new CD 16 using all of Summit. That makes it strong D (56.6%).

The above 5 districts are currently two strong D, two lean R, and one competitive R. My map makes them two strong D, one lean D, one competitive D, and one competitive R.

[muon2]

Georgia is actually fairly easy to do except for the Atlanta area counties. There are so many counties that none except the aforementioned are split IIRC.

How any large counties are internally divided affects the other counties (or parts) with which they will combine.

[muon2]

I have completed PA. There are three counties in excess of one CD population: Allegheny, Montgomery, and Philadelphia. I was able to make one whole district in Allegheny, two in Philadelphia, but none in Montco. I also need to split at least one other county, and a four way split of Lancaster avoids splitting anywhere else.

Within Philly I used the city planning districts as towns in the same way that townships, cities, and boroughs are used in the other counties. That way I preserved zero for rules P-1 2, and 3. Rule P-4 is one (Lancaster). Rule P-5 is one (Montgomery). Rule P-6 is 13 partial districts, and rule P-7 is four in Lancaster. Reducing the number of splits in Lancaster either increases P-4 or P-6.

Here is the map (you can tell which states I'm working on next. )

[muon2]

At this point I'm supposed to supply the polical analysis of the PA districts. I can do most of them, but I can't find election returns by town for Lancaster and Montgomery counties. Until then I can speculate about the overall district make up.

Based on the data I do have, there would be 11 R (seven strong, three lean, one competitive), 8 D (five strong, three competitive). The current delegation is 12 R, 7 D, but I don't have presidential votes per CD in 2004 to compare with the districts on my map. Again, if any of our many fans of PA have that data, I'll tabulate it.

[muon2]

I've added detail to the map above for NJ and upstate NY.

NJ: No towns needed to be split, rules P-1,2,3 are all zero. That's the good news. NJ has three couties large enough for a whole district, but another nine counties with over half the size of a CD. Splits become inevitable. For P-4, I split three (Burlington, Camden and Morris). For P-5 only one (Essex) while Bergen and Middlesex keep a whole district within. There are fifteen partial districts for P-6, and the maximum occurs many times: P-7 is three.

Upstate NY: NYC, LI, Westchester and Putnam equal 18 CDs. That leaves 11 for the rest of the state. For that area P-4 equals two (Oneida and Sullivan), and P-5 equals zero. With many counties, fewer partials are needed, P-6 is six and P-7 is two (Oneida and Sullivan).

[muon2]

Are New England seats all Dem execpt southern NH ?

Yes.

NH has one leans D seat and one competitive R seat based on the 2004 presidential vote. Using the 2000 vote they both would be competitive R.

ME has one strong D seat and one leans D seat with the map I posted. All the seats in MA are strong D in 2004.

ME