Concerning outliers

No, of course if the election were held today Barack Obama wouldn’t beat John McCain by fifteen points. A poll that shows a 55-36 partisan split may reflect a shift toward the Democrats, or it may reflect some sort of sample-selection bias; most likely it reflects some of each. And of course the election isn’t being held today.

Still, even outliers convey real information. Whatever the actual error is in polling (and it’s clearly much larger than the sampling error that gets reported as the plus-or-minus figure) it’s very unlikely that any competently done poll of a race that’s actually more or less tied yields a fifteen-point edge for one competitor. If the question is, “With Bush’s job performance ratings in the toilet, why is the Obama/McCain race still so close?” the answer may be “It isn’t.”

This is no time to be complacent. But it’s an excellent time to ignore the doomsayers. “Do not despair; one the thieves was saved. Do not presume; one of the thieves was damned.”