Sadly, Massimo, you seem to pay more attention to our politics than most Americans do!
In the spirit of fair trade, we would be happy to trade you and your family for some dead weight we’re carrying over here. :-D
Well, three of my biggest passions are politics, America and numbers.
I can combine the three of them here! :)
Anyway I made a mistake, my model had Missouri in (it was designed in 2008 when Missouri actually was a swing state) instead of Michigan. It changes quite a lot cause the first is a red state while the second is blue state.
Now I fixed it to have the exact same swing states Rasmussen included in his swing states tracking and so a 4 points lead in these 11 states combined would mean a 3.5 points lead nationally with Iowa and New Hampshire which would too go under the belt.
I mean this 4 points lead in the swing states tracking paints quite a more favorable picture than the single point lead in the national tracking.
Again, I can’t see how this can go together with such a tight race in Ohio.
I found a way to post the document on google docs.
So here it is
https://docs.google.com/file/d/0B2JIAz5DeQPtUUdDRlpMMG90OFU/edit
Quick explaination:
Sheet from 1 to 5 are just data acquisition. For each of the 11 states the excel sheet acquires a coefficient for every election for both parties.
For example in 2008 McCain in Florida had 48.22% while his national result was 45.70. So his result in Florida was 1.06 times “bigger” than his national figure, so 1.06 it’s the Gop coefficient in Florida 2008-—> KGop2008-Florida = 1.06.
You procede like this for each state for every election till you have a best coefficient and a worst coefficient for each state based on all the elections taken into account.
I created the model for the 2008 election and then I had 1996-2000 and 2004 elections in it. This year I dropped 1996 and put in 2008, cause behaviors of the states compared with the national average change overtime (slowly for some states, more quickly for others whit bigger demography changes) and I thought that 1996 data was not that significance anymore.
The goal is to have a model which tranforms the national result of each candidate in likely ranges (or averages) in each of the 11 swing states. So it doesn’t predict the national result, but once You input that in the model, transform it in results in the 11 states.
Of course this is not an exact method cause distributions are never exactly the same twice, but most of the time they change slowly enough to have a resemblance with their past :)
Results are in sheets 6 and 7.
Sheet 6 gives you ranges and average for each state based on 2000-2004-2008 distribution of vote.
Sheet 7 gives the result only based on the 2008 distribution.
The only input data, which can be changed every time one wants to, are the national result of each candidate in 2012 which are at the top of sheet 6 and 7.
In this case I have put Romney 50% and Obama 46.50 cause I was trying to determine what situation would have led the model close to a the swing states Resmussen scenario.
The 50-46 in the 11 swing states paints clearly a much more favorable picture for Romney than the 1 point lead in the presidential national tracking poll. Truth could be somewhere in between.