Anybody like this?
The idea is to create regions that are politically coherent -- sharing similar demographics and generally tending to vote together. And if you're playing at home, each region must have at least four states and be geographically contiguous, although you're allowed to cheat slightly on Alaska and Hawaii.
EDIT: Or this?
Friday, April 11, 2008
Regions
-- Nate Silver at 8:53 PM
Labels: cartography, meta, site
24 comments
That actually makes some measure of sense. I noticed a few states are grey -- does that indicate a lack of regional preference or is grey a preference itself?
Not sure if the Brown works that well(GA seems most out of place to me.), and I would think that PA belongs more with OH in blue then Yellow. Everything else looking pretty good.
The grey states are supposed to be their own region -- sort of the Moonshine / Appalachian-Ozark region. All of those states tend to be very hard to classify otherwise, and it seems to make some sense to group them together.
Yeah, Georgia sort of sticks out in that brown region. But it is changing, albeit slowly. In terms of education levels and incomes, it's starting to share much more in common with the brown states than the red states. It's also much more urban.
GA definitely belongs in brown rather than red. I lived there for many years.
IN has to be light blue. I can understand that culturally it belongs with the greys, but you can't have the great lakes without it. The only way you can leave IN where it is is if you make OH, MI and PA a region all on their own, being the three big swing states they are.
All the rest of it looks about right.
Speaking from Kentucky, I back your grey region. Kentucky has poverty like Alabama and Mississippi, but it has a small African-American population and a large Appalachian white population facing big economic challenges. Oklahoma might actually be an extension of the same culture, especially if you trace that Appalachian ethnic stock as it migrated.
Virginia, North Carolina, Georgia, and Florida have all been dramatically changed by immigration and urban development. South Carolina, though, seems to me like it belongs in the region with Alabama and Mississippi. (Non-Atlanta Georgia would fit there too, but greater Atlanta is now the main demographic story of Georgia.)
What would it look like at the county level (perhaps with a relaxed constraint on the 'contiguous' requirement there)
That's a good catch on Oklahoma, I think.
Indiana is the state that's nearly impossible to categorize. It really belongs with the orange states in terms of demographics.
Too many regions, in my opinion.
There's the Pacific Coast (blue), the West (red), the South (red), the Great Lakes (purple), Appalachia (purple), and the North-East (blue).
I personnaly would put together the "Deep South" States (LA, AL, MS, GA, SC) together, and NC and VA with the border states.
In my opinion, FL and TX are really closed socially, and more and more politically. So you can join them.
I agree with Sporcupine, Oklahoma belongs with KY, TN, WV, etc. Think of some of the demographics and the 2008 primary results. I also think IN belongs with these; I think of it as Kentucky with a bit of Chicago in it.
The others are about the best you can do; a Mid-Atlantic/rust belt category of MI, OH, PA, MD, DE, NJ, NY also makes sense from the numbers, although I think it would make more sense to have MD with VA.
No matter what you do something will look funny! And it will provide for endless debate... :)
Yeah, a county breakdown would be more telling. Virginia for example...its western counties voted in the primary much differently than the rest of the state, and is more of a forecast of the challenge Obama can expect to face in WV and KY. Florida sout of the panhandle is a group all it's own, and the panhandle would be red. And the Red River Valley of ND would fit in fine with the blue region next door.
I could go on and on!
I think that the regions you propose in both are interesting, the one thing I would suggest is that Virginia is becoming more "Mid-Atlantic" than "Tobacco South." It is unclear yet where it should be this election, but I think being grouped up with MD makes more sense than with NC, SC, and GA.
I like the first map the most out of the two, mostly because of your havint taking indiana out of the great lakes region, but i feel oklahomas placement is dubious, and arkansas could be really in either of the two southern regional categories you outlined.
The second map seems very good to me. Perhaps Indiana should be green though? I'm not sure it is fair to put them in the same toss-up status as MI, OH, PA.
Poblano,
Hi. I was curious about your views about the recently released general election polls for Pennsylvania and North Carolina by rasmussen at http://www.rasmussenreports.com/public_content/politics/election_20082/2008_presidential_election/pennsylvania/election_2008_pennsylvania_presidential_election
and
http://www.rasmussenreports.com/public_content/politics/election_20082/2008_presidential_election/north_carolina/election_2008_north_carolina_presidential_election
Have you added them yet to your simulations?
Thanks and take care
Hassan
I like the second regional map better than the first.
Here's their EV totals
Green - 34
Yellow - 62 (I added DC to this one)
Pink - 69
Light Blue - 48
Gold - 17
Purple - 18
Dark Blue - 77
Blue/Grey - 29
Red - 65
Green - 41
Brown - 80
Don't know why, but I felt like having the EV totals available.
Hassan, a quick look to the right could help. He added PA; but not NC. But I can help you there, at least for the State Win Percentage.
Obama (Popular Vote) -6,0
Clinton (Popular Vote) -10,0
Obama (Win%) 24,8%
Clinton (Win%) 11,5%
This should not differ too much from Poblanos numbers.
The first poll I had in my sheet before 538 had- a historic moment for me :)
Well, I still have to add NV-WY polls...
Rasmus, that's actually some movement in NC for Obama by almost +9% for the state
I don't think pablano will have that much movement. If it did though it should move the national win % up by about 1%
I think my left testicle is best put with the gray group and my right tesicle is more orange
bedir,
in my sheet his Win% moved up from 17,7 to 24,8.
This IS a difference, but the poll was taken by a good pollster, and it couldn´t be more recent.
In my sheet, that poll gets a weight of 1,09 (slightly different model, the regression gets 0,75, the other NC-poll 0,67/0,62/0,36/0,44/0,26 as listed on 538.com, I give more weight to the sample size and less to the pollster reliability. And I do not penalize polls when a new poll by the same pollster comes out.)
Ever ran your sim without the RegressioN?
I like the second one better. It flows better.
Finally 538´s Win% in NC for OBama moved up to 23,5%...
Not SO far away given the random noise that is- I don´t know why- biggest in the State Win%.
I think we would get out closer numbers when if simulated that let´s say... 25000x.
But that would take at least 20 minutes, not worth the time...