Dividing the US into regions

[Antonio the Sixth]

I was thinking about calculating the results of US Presidential elections by regions, and see if we could find consistent large-scale patterns. However, the main problem remains : how to define the US regions ? I'd like this thread to be a place to discuss about whta would be the most correct region breakdown.

First of all, here are a couple of possibilities I've studied.


"Big" regions

The first option is to draw very bold regional borders, like in this map, which is maybe the most basic :

Or in this one, which I personally prefer :

The latter is probably better because plain States( from ND to KS) has a lot more to do with places like MT and WY than with MN or MO and because the West Coast is a very different area than Mountain West. Still, both are pretty unperfect : the South's differences aren't visible, same for the West, which comprises both WY and NM.


A try : demographic regions

When compiling an excel spreadsheet about historical US populations, I wanted to regroup States with similar demographic patterns. The results were rather succesful, depite some possible mistakes :

NB Northeast and West Coast are 2 separate regions in this map, I just didn't have enough colors.


Small regions

This one is my most "intuitive" map, but other ones are to consider.

These are the regions used by the Census bureau.

And those are regions used by Nate Silver, which I initially didn't like but now realized that they really make some sense.

Finally, a couple of alternate modifications that could be done to the map (light shades mean hypothetical scenarioes) :

Expanding the "deep South" region to some more States and adding MD and DE to the Coast.

Creating a broad "appalachian" region with some hypothetical expansions in the South or the Rust Belt.

Creating a "ol' West" and a "new West" region (note that light States can be merged to become a third region).

Another, more North/South division.

So, now, I'd like you to discuss these few proposals, explain which make more sense to you and why, and possibly explore new possibilities.

[Antonio the Sixth]

Ok, there are several very interesting things there.

First of all, there is something to see with New England. I have the feeling that Libertas made a right point by separing Northern NE from Southern, and adding the latter to the Northeast. Politically, I also think Massachusetts is closer to NY than to Maine. On the other hand, a 3 States region is a bit small for these standards.

Also, I'm very annoyed by the hypothetical region going from MD to FL by not comprising SC and GA... I think it would make sense, but I'd try to avoid non contiguous regions. That's a true dilemma...

To Muon, I almost agree with your demographic map, but I'd rather put MN in the Great Lakes, WV in the Delta South and NV in the Southwest. I think I'm gonna edit my spreadsheet using those regions.

Then many other things to discuss, but it'd be far too long (for example, I'd tend to discard a West Coast including Nevada, a rust belt including Illinois, a two-States region in the Northwest corner or a Outer South with Louisiana, but all this is debatable). Now, to make some progress, let's try to start with the most basic possible map and then proceed by a State-by-State aproach. Here is the map that will allow us to begin :



So here are what I'd call "basic regions" : the groups of States, that, under almost every scenario, are associated together. Grey States are those that could fit in several regions. Note also that basic regions may be merged together : for example the light blue region mab be associated with the 40%-shade blue region or with the dark red one.

I'll go with the region's name so that we make no confusion (from the lightest shade to the darkest) :


North New England
South New England
Mid Atlantic
Southern West Coast

West Deep Sout
Deep South
Appalachia
Rust Belt

Midwest
Plains
Big Sky
Northwest

Southeast


Now, the main thing is to discuss the several possibilities for grey States, that's what I'll try to do now.

[Antonio the Sixth]

New York

- Merge it with PA and NJ, and possibly with Mid Atlantic.
- Merge it with South New England, NJ, and possibly Mid Atlantic.
- Merge it with SNE, NNE, NJ and MiAt, and possibly PA
- Merge it with SNE and NNE

New Jersey

- Merge it with NY, and MiAt, with or without NE and PA (but with at least one of those)
- Merge it with NY, PA and possibly MiAt
- Merge it with MiAt and SEC
- Merge it with NY, SNE and possiply MiAt

Pennsylvania

- Merge it with NY and NJ, possibly with NE (only the Sout or both) and/or MiAt
- Merge it with Rust Belt and Michigan
- Merge it with Rust Belt and Appalachia, with or without MI
- Merge it with NJ and MiAt

Michigan

- Merge it with Midwest, with or without MO
- Merge it with Rust Belt and possibly PA
- Merge it with RB, PA and Appalachia

Tennessee

- Merge it with Appalachia, AR and MO, with or without OK or Rust Belt (but not both)
- Merge it with the Deep South, possibly with WDS and AR (at least one of those)
- Merge it with DS, AR and TX, possibly OK
- Merge it with Appalachia and Rust Belt

Arkansas

- Merge it with Appalachia, TN and MO, with or without OK or Rust Belt (but not both)
- Merge it with DS, possibly with WDS and TN
- Merge it with DS, TN and TW, possibly OK

Missouri

- Merge it with Midwest, with or without MI
- Merge it with AR, TN, App, with possibly RB or OK (but not both)

Texas

- Merge it with DS and WDS
- Merge it with DS, AR, and possibly TN or WDS
- Merge it with DS, OK, AR and TN
- Merge it with Ok, Plains, Big Sky, Utah, AK and possibly CO
- Merge it to Southwest, NV, and possibly CO and UT

Oklahoma

- Merge it with AR, MO, TN and App
- Merge it with TX, AR, DS and possibly TN
- Merge it with Plains
- Merge it with TX, Plains, Big Sky, UT, AK and possibly CO

Colorado

- Merge it with Southwest, NV, and possibly UT
- Merge it with Southwest, NV and CA and possibly OR and WA
- Merge it with UT and BS, and possibly NV and AK
- Merge it with Plains, BS, UT, and possibly OK, TX, and AK

Utah

- Merge it with BS and possibly AK, CO or NV
- Merge it with BS, Plains, AK and possibly CO, OK and TX
- Merge it with CO, NV, SW and possibly TX

Nevada

- Merge it with SW, CO and possibly UT or TX
- Merge with CA, Northwest, HI and possibly SW
- Merge it with UT, CO and BS, and possibly AK

California

- Merge it with NW and HI
- Merge it with NW, HI, NV and possibly SW
- Merge it with NV, SW, CO and possibly TX or UT

I don't make AK or HI because they are pretty simple to place : either merged togehter, or with CA and NW, or with BS (for AK).

[Antonio the Sixth]

If you're ready to waste a few time, try to rank for each State the different possibilities beginning by those that make more sense. When I say "possibly" or "with or without", try to make your choice too.

[Antonio the Sixth]

Well, I certainly have a far lower knowledge of demographic data than you, so I am almost certainly right. I just noted that MN had a populaton growth pretty similar to Wisconsin, while IA, Mo and Plain States have seem a far higher Population erosion. As for Nevada, it's high population growth makes me think to Arizona, but it's true that the same is worth for CO and UT. Hence my reluctance to separate the "central West" (CO, UT, NV) from the southeast (AZ, NM, CA). Anywyas, we agree for most of the map, that's the most important.


Anyways, getting back to political regions, I've been trying to make some progress with our initial map, and came to a few conclusions. By doing so, I've reduced the number of pre-regions from 13 to 11 and the number of unallocated states from 15 to 8. Here is the result :



A few explanations :
- Alaska and Hawaii go to west coast and Big sky respectively, seems the most coherent solution to me.
- Northern NE is interesting in many ways, but making a region with only 3 States doesn't seem a good idea to me
- I've merged NY and NJ (which sense would it make otherwise ?). At this point, the only viable solution to make it a true region was to add MD and DE.
- Non-contiguous regions really bother me, so to avoid this scenatio I've merged EDS with SEC.
- I opted for two separate regions in the "new West", so to make the region viable I had to give it NV and CO.
- For the same reason, CA is incoroprated to the OR-WA region.

I'm somewhat desappointed, Devilman, because you were the only one answering me and I eventually came against almost all of your proposals. I'm obviously always open to discussion, but I also reflected a lot to that and I personally prefer these solutions.

Now, this map will highlight the possibilities for the remaining State. I admit for some States the result is pretty ugly, but I didn't know what else I could have dome besides stripes :



To put it more clearly, here are for each State the region it could be merged with :

- Pennsylvania Mid-Atlantic, Rust Belt
- Michigan Rust Belt, Midwest
- Tennessee Appalachia, Deep South
- Arkansas Appalachia, Deep South
- Missouri Appalachia, Midwest
- Oklahoma Deep South, Appalachia, Plains
- Texas Deep south, Plains, Southwest
- Utah Big Sky, Southwest

Also, an other thing : white lines mean that two regions might be merged together. Obviously all these choices are interdependent (for example it would make poor sense to give PA to Rust Belt if you merge it with midwest).

Now, tell me what would your choices be.

[Antonio the Sixth]

Think you need to check where the Mississippi river be before you put Arkansas into a region named after the Appalachians...

Oh, please...

1. The region name is the name of the region without other States merged. Appalachia=WV+KY.
2. The final region name will be determined once the regions themselves are determined.
3. Why do you care so much about region names ?
4. Since you saw this thread, what about making an useful contribution, by - I don't know... - posting a map or saying where you would put each State ?

[Antonio the Sixth]

Ok, it seems that this had made no progresses, so here is what my plan for last States. The choice to put one State in a particular region should be determined by several criteria. Let's list the few we can identify :
- I plan to make a measure the degree of electoral similarity between the States and each region, based on a handful of particularly representative Presidential elections. We will be able to see which region is more electorally "similar" to a particular state.
- Each region will be comprised of at least 4 States, and a maximum of, say 10. Area and population constraints should also be studied.
- I will open a poll for each state, asking you in which region you would put it. This will help me to take better your opinions into account, provided that enough people vote.
- Finally, I won't forget the general coherence of the final map, so it's important to think at the consequence for a region to put a state into another region.

So, what do you think about that ?

[Antonio the Sixth]

It would be easier if state lines could be altered...

Yeah, you're right, and that's why it was so easy for me to create regions in my "alternate US States" project with TX and CA both split in three States. Of course, those difficulties also make the thread more interesting, since more combinations are possible.

How can you create maps with all those colors?

Oh, that's quite simple. I used to think the Atlas maps have only the 4 colors that are presented in the EVCalc, but in realit there are 2 more : yellow and orange. To put it on your map, you have to use the map code.

For example, you have SS=x;y;z (S being a State's initials, x, y and z three numbers). The first number (x) represents the color. 0 represents grey (tie), 1=red (dem), 2=blue (rep), 3=green (ind), and then there are the two "hidden colors" 4 (yellow) and 5, which gives you orange when popular vote percentage are activated and white when they are not.

Example :

I've made MN yellow replacing MN=2;10;5 by MN=4;10;5, and NY white by replacing NY=2;31;6 by NY=5;31;6.

[Antonio the Sixth]

...To put it on your map, you have to use the map code.

For example, you have SS=x;y;z (S being a State's initials, x, y and z three numbers). The first number (x) represents the color. 0 represents grey (tie), 1=red (dem), 2=blue (rep), 3=green (ind), and then there are the two "hidden colors" 4 (yellow) and 5, which gives you orange when popular vote percentage are activated and white when they are not.

You seriously type all that stuff?  I could never stand to do that in a million years.

You don't have that much to type. There's a number written, which you have to replace with another number : not that complicated.

[Antonio the Sixth]

LOL

[Antonio the Sixth]

Ok, here are the polls' results :

Pennsylvania :
- Northeast : 12 (55%)
- Rust Belt : 10 (45%)

Michigan :
- Rust Belt : 20 (95%)
- Midwest : 1 (5%)

Tennessee :
- Outer South : 13 (76%)
- Deep South : 4 (24%)

Arkansas :
- Deep South : 11 (55%)
- Outer South : 9 (45%)

Missouri :
- Outer South : 10 (50%)
- Midwest : 10 (50%)

Oklahoma :
- Plains : 12 (67%)
- Deep South : 3 (17%)
- Outer South : 3 (17%)

Texas :
- Southwest : 11 (58%)
- Deep South : 8 (42%)

Utah :
- Big Sky : 12 (67%)
- Southwest : 6 (33%)

My votes aren't counted, and Vazdul reversed his call for MO and TX.

So, for 4 States we can see a clear majority (more than 2/3rds) emerging : Michigan for Rust Belt, Tennessee for Outer south, Oklahoma for Plains, and Utah for Big Sky. That would (I say would) give us this map :



Also, before we proceed to next step, I'd like to clarify something. Muon has evoked the possibility of putting Colorado in Big Sky (instead of Southwest). So, I thought I could make a poll also about it, but only if I see this idea is shared by enough people. Would someone else support Muon's proposal ?

[Antonio the Sixth]

The thing is, imo, that if you use more than say 5 regions the standards for accuracy you're developing really forces you to give up the concept altogether.

Seeing the difficulties and disagreements we're ahving there, I'm starting to think you aren't entirely wrong... However, I don't want to renounce and think it's worth to go until the end with this project.

[Antonio the Sixth]

Updated poll results :

Pennsylvania :
- Northeast : 15 (58%)
- Rust Belt : 11 (42%)

Michigan :
- Rust Belt : 22 (85%)
- Midwest : 4 (15%)

Tennessee :
- Outer South : 14 (70%)
- Deep South : 6 (30%)

Arkansas :
- Deep South : 11 (50%)
- Outer South : 11 (50%)

Missouri :
- Midwest : 13 (54%)
- Outer South : 11 (46%)

Oklahoma :
- Plains : 17 (68%)
- Deep South : 4 (16%)
- Outer South : 4 (16%)

Texas :
- Southwest : 13 (57%)
- Deep South : 10 (43%)

Utah :
- Big Sky : 15 (68%)
- Southwest : 7 (32%)

Threee vote changes taken into account : vazdul for Missouri and Texas, Opebo for Tennessee.

3 majorities have expanded : Pennsylvania, Oklahoma and Utah. 3 have narrowed : Michigan, Tennessee and Texas. In Missouri, a tie has turned into a narrow Midwest lead and in Arkansas, a narrow Deep South lead has become a tie.

Of course, I encourage those who haven't yet voted to express themselves and vote. those polls aren't supposed to close, and if I see major changes I will update the results again.

[Antonio the Sixth]

Also, since nobdy seems to support transfering Colorado into Big Sky, I definitely place it it in the Southwest.

Ok, now time is up for the second step : studies of Presidential election results. First of all, what I want to do now is to establish a panel of 10 elections which are the most representative of regional divides in the US. For obvious reasons, it wouldn't make much sense to look for them before the end of the reconstruction, when these regions made absolutely no sense. So, here is the challenge : from 1876 to 2008, we have to find the 10 elections that will most help us to establish regional patterns.

The first elections I managed to isolate were those :
- 1884 (the most perfect example of the regional polarization of the USA in an extremely close election)
- 1896 (major regional realignment)
- 1912 (races with a great number of strong candidates are always useful to analyze)
- 1924 (same reason)
- 1928 (an important election, with Smith deeply changing the nature of Dem electorate)
- 1932 (New Deal realignment)
- 1964 (upset reversion of the Deep South)
- 1968 (same reason as 1912 and 1924)
- 1972 (Nixon sweeping the South by enormous margins)
- 1976 (East-West divide)
- 1984 (establishment of the new Republican coalition)
- 1992 (Clinton managing to rally the Outer South while the Deep south remains republican)
- 2000 (very polarized election)
- 2008 (the most recent one)

The problem is : here we have 14 elections, how do we manage to reduce this number to 10 (or maybe even less) ? I initially thought to simply exclude 4 of them, but eventually realized they were all useful in some way. that's why I thought the best thing to do was to establish a panel adapted to each State. Starting with those 14 elections, we onl keep those that show something useful for a particular State (or a particular couple of States). Now, let's get back to our 8 states we want to classify. Which group of elections you think would best fit for each of them ? Let's discuss there.

[Antonio the Sixth]

Thanks a lot Vazdul. This analysis already allow us to draw a closer tie between Tennesse and Outer South, Utah and Big sky and between Oklahoma and Plains (those two can already be considered as assigned), as well as, more surprisingly, between Missouri and Outer South, which goes against our previous poll. Also, this also draws a clear relation between PA and NE and between MI and RB. No clear pattern can instead be found for Arkansas and Texas.

Now, to expand this research a little bit, I'll add some elections to thos Vazdul has mentioned, in order to compose my "panel". As I've said, will be made of 10 elections.

Pennsylvania : 1884, 1896, 1912, 1928, 1932, 1968, 1976, 1984, 2000, 2008
Michigan : 1896, 1912, 1924, 1928, 1932, 1968, 1976, 1984, 2000, 2008
Tennessee : 1884, 1896, 1928, 1932, 1964, 1968, 1972, 1984, 1992, 2008
Arkansas : 1884, 1896, 1928, 1932, 1964, 1968, 1972, 1984, 2000, 2008
Missouri : 1884, 1896, 1912, 1932, 1968, 1972, 1976, 1992, 2000, 2008
Oklahoma : 1912, 1924, 1932, 1964, 1968, 1972, 1976, 1992, 2000, 2008
Texas : 1884, 1896, 1928, 1932, 1968, 1972, 1976, 1992, 2000, 2008
Utah : 1896, 1912, 1928, 1932, 1964, 1968, 1984, 1992, 2000, 2008

[Antonio the Sixth]

Well, the in-depth results comparison will probably tell us more about that.

[Antonio the Sixth]

Well, the in-depth results comparison will probably tell us more about that.

I get the wink, but I'm not sure how you'll bring this to conclusion other than to go with your gut. When you polled the forum, I would have thought the natives of the states should get extra weight - such as opebo's comments on MO. If the goal is to understand trends in voting patterns at a regional level, then there are many factors, both analytical as well as perceptual. Use of only superficial factors will tend to produce superficial conclusions.

I don't inted to use the electoral similarity as a bible and follow anything that will come out. Like the polls, they are mere hints. I'll take into account the polls, the election data, and several ofther factors to take my decision. Of course, your point of view is very useful, as well as those of some other users like Vazdul. But it doesn't mean the rest becomes useless.

And well, there are usually not more than 1 or two users residing in a polled stete, so I can look at their comments but not really weigh their vote differently.

[Antonio the Sixth]

Oh, I'm sorry. I've just started University so I didn't have as much time as I used to when I started this. Of course I'm still very interested to this project and will update it as soon as possible. Tomorrow I'll try to use with my comparison algorithm for Pennsylvania, though I can't promise you I will.

[Antonio the Sixth]

Cute, I was writing a post and my computer tilted !

Anywyas, I was saying that I'll start my state-by-state comparison for Pennsylvania (comparing with the States that compose or coulpd compose the region in which Pennsylvania may be included : ie NY, NJ, DE, MD, DC, OH, IN, MI, WV, KY and TN). So I was saying that to do so I will compare the percentages got by each candidate, each time I'll take the lower of the two, and I will sum all this. The more the result is close to 100, the more the states are similar. the more it's close to 0, the more they're different. I'm gonna call this index "similarity rate". For example, I'll calculate similarity rate for PA and NY in the 1884 election :

Candidate	PA	NY
Cleveland	43.46%	48.25%
Blaine	52.97%	48.15%
St. John	1.69%	2.14%
Butler	1.88%	1.46%
Others	0%	0%

So, the similarity rate is 43.46+48.15+1.69+1.46 = 94.76

This is the method I will use for next comparisons, so wait and see.

[Antonio the Sixth]

Now let's start seriously.

Pennsylvania

1884

- Ohio : 96,52%
- New York : 94.76%
- Indiana : 94%
- New Jersey : 93.8%
- Michigan : 93.36%
- West Virginia : 92.52%
- Tennessee : 92.01%
- Maryland : 91.39%
- Kentucky : 88.14%
- Delaware : 86.9%


1896

- New Jersey : 98.45%
- New York : 96.6%
- Maryland : 93.75%
- Michigan : 92.79%
- Delaware : 92.2%
- West Virginia : 89.44%
- Ohio : 89.2%
- Indiana : 88.31%
- Kentucky : 87.41%
- Tennessee : 84.17%


So, what can we conclude from thos first two elections ? For 1884, not so much : Rust Belt States and Northeast States have more or less the same level of similarity (though Maryland and Delaware  are really low), so I rate this 1st election Inconclusive. As for 1896, the similarity is clearly superior for Northeast, so I rate it Pro-Northeast.

[Antonio the Sixth]

Also, here's some update of polls (which BTW are still going on, so new voters would be welcome ).

Pennsylvania :
- Northeast : 16 (57%)
- Rust Belt : 12 (43%)

Michigan :
- Rust Belt : 24 (86%)
- Midwest : 4 (14%)

Tennessee :
- Outer South : 16 (73%)
- Deep South : 6 (27%)

Arkansas :
- Deep South : 12 (52%)
- Outer South : 11 (48%)

Missouri :
- Midwest : 15 (58%)
- Outer South : 11 (42%)

Oklahoma :
- Plains : 19 (68%)
- Outer South : 5 (18%)
- Deep South : 4 (14%)

Texas :
- Southwest : 13 (52%)
- Deep South : 12 (48%)

Utah :
- Big Sky : 17 (71%)
- Southwest : 7 (29%)

Threee vote changes taken into account : Vazdul for Missouri and Texas, Opebo for Tennessee.

Majorities increased in Michigan, Tennessee, Missouri and Utah. They narrowed in Pennsylvania, Oklahoma and Texas. In Arkansas, the tie has become a narrow Deep South lead.

[Antonio the Sixth]

Oh boy, talk about a blast from the past. This was back in the days when my command of English was still a little shaky. Fun times.

Anyway, I've long renounced trying to break up the US into regions based on States. The only way to do it right would be to go down to the county level. I think Muon had a solid attempt in this regard (though I'm not sure I agree with all his choices).

[Antonio the Sixth]

Oh boy, talk about a blast from the past. This was back in the days when my command of English was still a little shaky. Fun times.

Anyway, I've long renounced trying to break up the US into regions based on States. The only way to do it right would be to go down to the county level. I think Muon had a solid attempt in this regard (though I'm not sure I agree with all his choices).

Link to muons thread?

I don't remember what it's called, but you can PM him. Actually, if he still has it, I'd be grateful if he could bump it so I can keep a record of it.

[Antonio the Sixth]

Thanks!!