2012 county results - Uniform swing by state now calculated

[Averroës Nix]

Yeah, the number of voters who live in counties with populations under 10,000 is almost negligible. I tried to use your exact method and it wasn't revealing anything interesting so I scrapped it.

The map that I posted applied a fifteen point swing to counties with fewer than 10,000 voters, a 2.5 point swing to counties with between 10,000 and 100,000 voters, and a 2 point swing to counties with over 100,000 voters.

Maybe if I have time later I'll calculate county population densities and play with that. Why do you expect such a huge increase in the disparity between urban and rural voting patterns?

[Wonkish1]

Yeah, the number of voters who live in counties with populations under 10,000 is almost negligible. I tried to use your exact method and it wasn't revealing anything interesting so I scrapped it.

The map that I posted applied a fifteen point swing to counties with fewer than 10,000 voters, a 2.5 point swing to counties with between 10,000 and 100,000 voters, and a 2 point swing to counties with over 100,000 voters.

Maybe if I have time later I'll calculate county population densities and play with that. Why do you expect such a huge increase in the disparity between urban and rural voting patterns?

A) I've read polling data that points to rural percentages being the largest swing in the electorate since 2008 by a huge margin.
B) A large degree of anecdotal evidence I've accumulated echo's the same thing in regards to the degree of change in people(and no I don't live in rural America).


Basically, the initial numbers you showed were a 2.5 move against Obama and a 3 pt increase in Romney's, right?

Well take whatever assumption you want to use and set up a way to give rural counties a huge move and medium sized counties less, and suburban/city counties even less move.

Then post the map.   I trust you to figure out some way to do that where the assumptions seem about right.

But for example in the state that I grew up in:
-Chippewa will be for sure GOP
-Clark will be for sure GOP
-Marathon will be for sure GOP
-Lincoln will be for sure GOP
-Langlade will be fore sure GOP
-Oconto will be for sure GOP
-Outagamie county will be for sure GOP
-Brown county will be for sure GOP
-Shawano county will be for sure GOP
-Waupaca county will be for sure GOP
-Mantiwoc county will be for sure GOP
-Racine county will be for sure GOP


Being overly generous to Obama the swing counties would still be: Door, Kewaunee, Marinette, Dunn, Winnebago, etc.

[nclib]

3 pt swing?  I think it will be a larger swing, but whatever. It's 538.

If you think that a larger swing is probable, give me a number for each candidate and I'll map it. It doesn't take long.

Is it possible to control for separate swings by state, according to 538's state predictions?

[Okay, maybe Mike Johnson is a competent parliamentarian.]

Sorry if this would be too time-consuming, but could you do a map based on the projected swings in the several states on 538? They have for example a 0.5% swing towards Obama in Arizona, 2.6% away from him in Ohio, 4.8% away in Illinois, 7.9% away in Massachusetts...

[Averroës Nix]

It wouldn't take terribly long - I'll see what I can do. Are Silver's predicted margins listed anywhere besides the cumbersome side panel on the blog's main page?

[Okay, maybe Mike Johnson is a competent parliamentarian.]

I'm afraid the cumbersome side panel seems to be what we have to go by.

[RI]

I can make some too.

[Badger]

There's no such thing as a Reagan Democrat.

At least not in SWPA--Carter and Mondale both swept the region in 1980 and 1984.

If one interprets the term "Reagan Democrat" to literally mean Democrats who liked and supported Reagan, then yes it's a weak analogy considering even Mondale strongly outperformed Carter in the region due to the utter devastation the steel industry and manufacturing in general suffered there during Reagan's first term.

If, however, one takes the term as shorthand for culturally and socially conservative blue collar whites who will vote Republican over gun control, abortion, and foreign policy, plus hold underlying suspicion towards the national Democratic Party over race and welfare, then yeah, the region is chock full of them.

[Bacon King]

Aye, I was going to recommend 538's predicted state swings to make a map, which would be a lot more accurate (especially for states like Wisconsin).

I don't think using county population alone to make a swing would really do what Wonkish is trying, by the way. A much more useful measure would be county population density, because a lot of rural counties are really big, but sparsely populated: compare, for example, Cocino County AZ with the nearby Washington County UT. They both have a population of ~135k but the latter is eight times denser. Percent urban/rural would probably be even better, actually, if such statistics are easy enough to find.

Of course Wonkish's method should probably also control for race in some way, so it doesn't cause unrealistic huge swings in the black belt, on native reservations, rural hispanic areas, etc. But at that point it just sounds way too complicated. I imagine that a uniform swing by state is the best option, because the only thing that would really offer a more accurate map would be probably way too complicated.

[Wonkish1]

Aye, I was going to recommend 538's predicted state swings to make a map, which would be a lot more accurate (especially for states like Wisconsin).

I don't think using county population alone to make a swing would really do what Wonkish is trying, by the way. A much more useful measure would be county population density, because a lot of rural counties are really big, but sparsely populated: compare, for example, Cocino County AZ with the nearby Washington County UT. They both have a population of ~135k but the latter is eight times denser. Percent urban/rural would probably be even better, actually, if such statistics are easy enough to find.

Of course Wonkish's method should probably also control for race in some way, so it doesn't cause unrealistic huge swings in the black belt, on native reservations, rural hispanic areas, etc. But at that point it just sounds way too complicated. I imagine that a uniform swing by state is the best option, because the only thing that would really offer a more accurate map would be probably way too complicated.

I figured that what you're suggesting was too cumbersome to do so I simplified.

I took it that the goal was to arrive at some approximation of the overall color of a county by county map. It was for the sake of simplicity that I recommended going off of population even though it would produce some odd results in various more Democratic rural counties around the country.

Hell if you wanted to get technical, take the 2004 and 2010 county by county around the country and for every county with a low population density(lets say defined as the bottom 70% density wise) apply the higher performance for the GOP of those 2 years(have to because many 2010 races went unopposed, etc.) and then take the balance of performance shift and apply it to the higher density counties so that the average spread assumption stays the same.

Now that is about the best you're going to be able to do, but even that isn't going to be that close either.

[RI]

Here's a map I made based on the polling averages by state since Oct 15 (or the latest if none have been conducted since then). Each state had a uniform swing within it, so I don't actually think this is what the map will look like exactly.

[Okay, maybe Mike Johnson is a competent parliamentarian.]

Yeah, even within states uniform swings are distinctly likelier in some places than in others. Obama is likely to outperform 2008 in southwestern Pennsylvania, for instance.

[Averroës Nix]

Sorry if this would be too time-consuming, but could you do a map based on the projected swings in the several states on 538? They have for example a 0.5% swing towards Obama in Arizona, 2.6% away from him in Ohio, 4.8% away in Illinois, 7.9% away in Massachusetts...

[Vosem]

Could you do the 538 map, but add +1.2 to Romney's figures and subtract -1.2 from Obama's figures? Adds up to a swing of 2.4 (swinging CO, IA, NH, OH, and VA). That might be interesting.

[Averroës Nix]

I don't think using county population alone to make a swing would really do what Wonkish is trying, by the way. A much more useful measure would be county population density, because a lot of rural counties are really big, but sparsely populated: compare, for example, Cocino County AZ with the nearby Washington County UT. They both have a population of ~135k but the latter is eight times denser. Percent urban/rural would probably be even better, actually, if such statistics are easy enough to find.

It is - acquiring usable data from the Census, ACS, BLS, or whatever is usually easy. I'm less sure of what I could do with it. Same with county-level data on race and ethnicity, or virtually anything else (income, change in home values, foreclosure rates, education levels, unemployment rates, etc.). I have my own intuitions about how each of these factors will affect the results of the election, but I'm not sure of how to build those into a quantitative model in a way that's likely to increase its predictive power.

Vosem, I'll post the map you requested in a moment.

[Averroës Nix]

Could you do the 538 map, but add +1.2 to Romney's figures and subtract -1.2 from Obama's figures? Adds up to a swing of 2.4 (swinging CO, IA, NH, OH, and VA). That might be interesting.

[Vosem]

Thank you very much, Averroes. It's amazing how contiguous GOP-voting territory is; you can go from the Pacific coast to Putnam County, NY, without ever stepping into a Democratic county.

[Wonkish1]

Could you do the 538 map, but add +1.2 to Romney's figures and subtract -1.2 from Obama's figures? Adds up to a swing of 2.4 (swinging CO, IA, NH, OH, and VA). That might be interesting.

That's about right. You're looking at a razor thin map for slight Romney right there.

For example:
-Flip Larimer county and make Jefferson almost a pure tie and Colorado probably flips to Obama in that map.
-Flip Dunn and Kenosha counties and Romney takes Wisconsin in that map.
-Flip Richland and Sandusky counties and Obama takes Ohio.