Traditionally, statistical significance in conclusion A means 95% confidence in conclusion A. The frequentist interpretation of "95% confidence" is that if twenty studies respectively yield conclusions A, B, C, D,..., Q, R, S, T each with 95% confidence, then we expect exactly one of these conclusions to be wrong. (It doesn't matter whether these studies repeat the same experiment or are about completely different subjects.) With that in mind, Nate Silver at FiveThirtyEight currently projects Obama will win the electoral college with 95.7% confidence. Silver kindly offers McCain advice on optimizing his slim chance of winning. (1. Abandon Pennsylvania, Wisconsin, Minnesota, Iowa. 2. Attack New Hampshire and New Mexico. 3. Defend Colorado, Virginia, Nevada, Ohio and North Carolina. 4. Gamble on Florida, Missouri and Indiana.) However, Silver is confident McCain will lose, and so am I.

I'm not happy about this situation. I agree with Krauthammer's endorsement of McCain for foreign policy reasons, except I'm less sanguine about the economy than Krauthammer is. It is true that, "Today's economic crisis, like every other in our history, will in time pass." It's also true that in 1933, a rhetorically gifted President and copartisan supermajorities in congress were elected after an economic crisis. They then greatly expanded the size of government and inadvertently made the economy worse. I don't think things will be nearly that bad this time around, but I still would greatly prefer a divided government in 2009.