Thanks, but now the spreadsheet in the OP doesn't make much sense. The whole idea of the elasticity was to determine how much the state would shift when the national average shifted. The spreadsheet applied the elasticity directly to the margin which is not how it's designed to be used.
For example, Utah has a -48.04% margin in the first row of the spreadsheet. I'm not quite sure where that's from since the Obama-Romney margin was -47.88% according to Atlas, but I'll take it as the 2012 margin as described in the post. If I then consider a 10 point swing to the democrats, one might expect the elasticity of 1.01 to generate a +10.1% change in the margin, but instead the spreadsheet just divided 48.04 by 1.01 which doesn't really mean anything.
Also, as I read Silver's article I see that he also recognizes that the elasticity would be greatest for a 50-50 state. A straight swing would be the dashed line in his graph above, but the real swing would be better modeled by the red curve. The actual shift described in my example above would be less than 10.1%. A proper model would be to look at the difference in the red curve from the black line compared to their difference at 50% and use the mirror image curve for a negative shock.
A simple way to approximate this with a spreadsheet is to create an elasticity weight equal to the difference between the margin and 100%. For UT the weight would be 51.96%. The find the expected shift, multiply the national shift by the elasticity and the weight. Thus a 10% national shift for the Dems would reduce the margin in UT from -48.04% to -42.79%.
