Using Nate Silver's "Swing Voter" Analysis to Determine Target States

[crypto-fascist superhero]

So Nate Silver just released this analysis of how "swingy" each state's voters are, that is to say whether a national swing of 1 point causes a swing in that state of more or less than 1 point.  I find this a very legitimate concept and have been sort of searching for a way to measure it myself, but of course ninja'd.  I did, however, calculate how difficult it would be for each party to swing each state from it's 2008 result, using the formula (margin/Silver-swinginess) to determine the national swing that would swing that state.  I would say that, looking at the list, it appears to be quite accurate by my own estimation.  Each state is listed with the national swing required from the 2008 result and the 2008 result if adjusted to be a 50-50 election.

Republican target priority:

North Carolina: 0.36/-6.90
Indiana: 1.04/-6.22
Florida: 2.68/-4.58
Ohio: 4.28/-2.98
Virginia: 6.63/-0.63
New Hampshire: 7.51/0.25 (wins election)
Colorado: 7.72/0.46
Iowa: 8.15/0.89
Minnesota: 10.14/2.88
Pennsylvania: 11.09/3.83
Nevada: 11.56/4.30
Wisconsin: 12.64/5.38
New Mexico: 13.16/5.90
Maine: 13.64/6.38
Oregon: 14.09/6.83
New Jersey: 14.65/7.39
Washington: 15.12/7.86
Michigan: 15.96/8.70
Connecticut: 21.10/13.84
Rhode Island: 21.56/14.30
Massachusetts: 21.69/14.43
California: 23.56/16.30
Illinois: 25.11/17.85
Delaware: 25.49/18.23
Maryland: 28.27/21.01
New York: 29.84/22.58
Vermont: 30.34/23.08
Hawaii: 36.50/29.24
DC: 190.93/183.67 (i.e. impossible)

Democratic target priority:

Missouri: 0.13/7.39
Montana: 2.13/9.39
Georgia: 6.12/13.38
North Dakota: 7.39/14.65
Arizona: 7.50/14.76
South Dakota: 7.65/14.91
West Virginia: 11.38/18.64
Texas: 11.76/19.02
South Carolina: 12.47/19.73
Kansas: 13.94/21.20
Nebraska: 14.64/22.70
Tennessee: 14.91/22.97
Kentucky: 16.06/23.32
Alaska: 18.10/25.36
Arkansas: 19.65/26.91
Mississippi: 20.90/28.16
Louisiana: 23.58/30.84
Idaho: 23.87/31.13
Utah: 27.74/35.00
Wyoming: 31.30/38.56
Alabama: 32.21/39.47
Oklahoma: 33.65/40.91

[shua]

http://graphics8.nytimes.com/images/2012/05/21/us/politics/fivethirtyeight-0521-elastic2/fivethirtyeight-0521-elastic2-blog480.png

I'm surprised Indiana is sort of low here given the huge swing from 2004 to 2008.

How are you getting the number in your denominator?

[crypto-fascist superhero]

It's the number in that chart.

[Gustaf]

That's pretty cool, I've been thinking about that issue myself.

[Gustaf]

Providing the obligatory maps: here is one with the states Silver calls swing states in grey:



That's GOP ahead of DEM 191-186.

Here is one where I assigned states that are inelastic or have average elasticity to where they would likely fall:



Throughout I assume 50-50 election, roughly speaking. That's a 235-232 map for Romney.

So that leaves Nevada and Ohio as states of average elasticity that are fairly close. And Virginia as a state with low elasticity which is close. And the rest as high elasticity states.

[phk]

I have done this before, but only from comparing 1996-2000 (two-way), 2000-2004,2004-2008.

[shua]

It's the number in that chart.

I'm a bit confused. I don't see any negative numbers on that chart or anything greater than 1.29

[Senator Napoleon]

It's the number in that chart.

I'm a bit confused. I don't see any negative numbers on that chart or anything greater than 1.29

The negative numbers would be for the 2012 margin. Romney could lose by nearly seven points and still pick up NC is the idea.

[Nichlemn]

I imagine the relationship starts to break down once you get into mega landslide territory. Obviously a Republican winning 100% of the vote is winning DC.

[crypto-fascist superhero]

I imagine the relationship starts to break down once you get into mega landslide territory. Obviously a Republican winning 100% of the vote is winning DC.

Well yeah but that's pretty much impossible by definition ... I'd say it probably works well up to about a 35-point victory, which is probably about the theoretical max anyway.

A 29.02% Democratic swing relative to 2008 (or a 36.28 point victory) would have the Republican getting less than 0% in DC.  I'd say 35 points it is.

[phk]

This is kind of cheating,  but the only way to really do this is to just have pollsters ask the questions "Are you a committed D voter?" and "Are you a committed R voter?", sum it up and subtract from 100.

[Nym90]

Interesting, though the idea that Indiana is an "inelastic" state seems a bit dubious considering that it actually had the largest swing from 2004 to 2008 of any state other than Hawaii.

[crypto-fascist superhero]

This is kind of cheating,  but the only way to really do this is to just have pollsters ask the questions "Are you a committed D voter?" and "Are you a committed R voter?", sum it up and subtract from 100.

The problem with that is that people lie and say they're "independent," "moderate," and/or "undecided" when they are in fact strong partisans.

Interesting, though the idea that Indiana is an "inelastic" state seems a bit dubious considering that it actually had the largest swing from 2004 to 2008 of any state other than Hawaii.

Silver said he made a distinction between voters actually changing their minds (which is what this measures) vs. simply increasing turnout for one's own side.  I'm guessing the latter was the case with Indiana.