I have run a state-wide polling operation for a Democratic president so I've learned a little about polling. Recently I've written about the statistical phenomenon Regression to the Mean Sarah Palin and regression to the mean, and why it predicts swings in election polling. Because of this phenomenon, I firmly believe Obama will win [prediction: 52-46], because the MEAN [the winning number in this case] will result in the majority of the electorate voting their concerns on the economy, not on personalities, as McCain is hoping for.
Yet I've noticed the Rasmussen pollsters edging more toward McCain in the last weeks, and indeed, throughout this election, they've been weighed heavily against Obama, with Clinton as his opponent and now with McCain. Rasmussen explains his rightward drift by stating that the electorate is becoming more Republican, particularly as a result of the Palin pick, and that the GOP vote is more enthusiastic, closing the gap with Obama voters.
This explanation explains NOTHING. It is a weak, facile, unsupported argument looking for a favorable result.
Perhaps the GOP share of the electorate has increased, but Rasmussen's numbers don't prove that. All of this 'increase' can be explained by interpreting the increased enthusiasm of GOP voters as an increase in numbers. In other words, if 30% of McCain's voters were unenthusiastic before the Palin choice and now that number has decreased to 5%, Rasmussen seems to be extrapolating that that denotes an increase in the number of GOP voters in the entire electorate. But at best, it is a guess.
And is it even a good guess? Would anyone here argue that Obama voters are less enthusiastic today than they were 3 weeks ago--I see no evidence of it. I do see some enthusiasm, or at least curiosity, since Palin, but she doesn't seem to have changed any Obama voter's mind. But according to Rasmussen, the GOP increase largely comes out of the Democratic Party's share. My theory is, Rasmussen is weighing his sample by his perceived enthusiasm screen. And that might be a valid method, but more likely it weighs too heavily toward McCain, because his base line enthusiasm was so low prior to Palin.
There is NO statistical evidence to suggest that Palin has caused voters to shift parties. What we DO know is that Democratic registration is up far in excess of Republican registration. Nationwide. In 2004, where the GOP registered voters in comparable numbers to Democrats, bush held his own. In 2008 there is NO state where the GOP is matching Democratic registration.
Does Rasmussen use registration as a weight in his likely voter model? No. In fact, he discounts for new voters. They do not exist in his polling universe. But voters that now express enthusiasm for a candidate they couldn't stomach up till now..their vote is worth x+ some coefficient known only to Rasmussen.
If you look at the election from the perspective of the pollsters, America is ready to give bush his 3rd term. They really aren't concerned about the economy. The enthusiasm that brought record high Democratic turnout to the polls this year, has dissipated.
I don't buy it. By my reckoning, the MEAN is still far left on the bell curve from where it was in 2004, and that favors Obama. Perhaps this is what Bill Clinton knew when he suggested this week that Obama will win handily.
The Country is going to be surprised this year when the election results come in. McCain will lose all Kerry states plus NV, NM, CO, IO, and probably Ohio and Indiana. Indiana?? you ask! That is my canary in the coalmine. Keep your eye on Indiana. Obama knows he can win it. He's not there yet, but the momentum is building.
And one last thing: I spend most of my time studying economic charts. I'm convinced that there WILL BE an October surprise, and it WILL BE an economic surprise, and it WILL NOT be good for McCain.