“To say that I’m unhappy would be an understatement,” said Harry Wilson, head of the polling center at Virginia’s Roanoke College, which projected 5-point Virginia wins for both Romney and GOP Senate candidate George Allen on Oct. 31. Romney lost the state by 3 percentage points, and Allen lost by 5. “I was drinking that Republican Kool-Aid,” Wilson said.

Wilson and other pollsters cited the same miscalculation: an assumption that the electorate that showed up Nov. 6 would be older, whiter and more Republican than the one that actually turned out. Obama’s victory in 2008, juiced by higher-than-normal turnout by young voters and minorities, was seen as an aberration, unlikely to be repeated in a struggling economy…

Paleologos said his decision to stop polling Florida, Virginia and North Carolina was based on limited resources and a desire to poll Colorado and Ohio, but also on “the incumbency rule,” a long-accepted premise that if an incumbent cannot rise above 47 percent or so in head-to-head polling, he is unlikely to win.

“What I’ve learned is that there’s a new norm,” he said. “The incumbency rule does not hold, at least in Florida and Virginia.”