It's not impossible. I roughly reversed engineered Silver's tipping point simulation... - Edit 2
Before modification by Burr at 05/11/2012 11:09:49 PM
His simulation is more nuanced, but I got mine to where it was giving similar results. My model focused on the notion of a "deciding state," and my definition of a "deciding state" is slightly different than Silver's definition of "tipping-point state." My definition of a deciding state was a state that a) had enough electoral votes to tip the election, b) would have flipped the election had the loser won it, and c) was the most efficient of those states in terms of electoral votes and the candidate's likelihood of flipping that state. My results had Florida as more likely to be a "deciding state" than Wisconsin, whereas Silver's had Wisconsin as more likely to be a "tipping point" state than Florida. Other than that, our order was pretty much the same for all the most important states in this election.
Once I got that bit working, I went back into the formulas and adjusted my measure of efficiency for state population. So it was then sort of a combination of a deciding state model and a return on investment model. With all else being equal, it is more efficient to pour resources into a state with fewer individuals than a state with more individuals, because the smaller state's voters are worth more individually than the other's.
I refer to the new model as my Facepalm model, because in a sense it measure the likelihood that a candidate will hit himself if he loses the state.
Having done that, suddenly Pennsylvania jumped up to be second in importance to Ohio.
However, one thing my new model doesn't account for is saturation. Ohio has already been saturated with political ads for weeks. So even if it outranked Ohio, Pennsylvania still might well be the more truly efficient state to pour ad resources into. And it remains very important for turnout resources.
But that is all just a long way of saying, very recently Pennsylvania was a major Facepalm state. Now, granted, Obama's poll advantage in the state has improved since his most recent dip. And I've been too busy to run my simulation again with the new numbers. (It's been about a week, I think.) Nevertheless, I doubt things have changed so much that Pennsylvania isn't still a significant Facepalm state for Obama. He'd better watch out there, or he might find Romney's been pouring ad and turnout resources into the just the right parts of the state to win it.
Once I got that bit working, I went back into the formulas and adjusted my measure of efficiency for state population. So it was then sort of a combination of a deciding state model and a return on investment model. With all else being equal, it is more efficient to pour resources into a state with fewer individuals than a state with more individuals, because the smaller state's voters are worth more individually than the other's.
I refer to the new model as my Facepalm model, because in a sense it measure the likelihood that a candidate will hit himself if he loses the state.
Having done that, suddenly Pennsylvania jumped up to be second in importance to Ohio.
However, one thing my new model doesn't account for is saturation. Ohio has already been saturated with political ads for weeks. So even if it outranked Ohio, Pennsylvania still might well be the more truly efficient state to pour ad resources into. And it remains very important for turnout resources.
But that is all just a long way of saying, very recently Pennsylvania was a major Facepalm state. Now, granted, Obama's poll advantage in the state has improved since his most recent dip. And I've been too busy to run my simulation again with the new numbers. (It's been about a week, I think.) Nevertheless, I doubt things have changed so much that Pennsylvania isn't still a significant Facepalm state for Obama. He'd better watch out there, or he might find Romney's been pouring ad and turnout resources into the just the right parts of the state to win it.