Some unexpected results in Tuesday night's four high-profile elections raise the question of which pollsters got the races right and which got them wrong.
The races that tripped up the polling community the most were the two big ones in New York. Not a single poll had Democrat Bill Owens, who won his congressional race in New York's 23rd district, garnering more than 37 percent of the vote, according to Pollster.com. He ended up with 49 percent.
Siena Poll came close to nailing the support behind the Conservative Party's Doug Hoffman -- putting his backing at 41 percent, four points shy of his actual result. But the same pollster drastically missed Owens' backing -- putting him at 36 percent. Public Policy Polling declared "Hoffman primed for dominant victory," and it projected him winning 51 percent of the electorate, with 34 percent for Owens.
The problem, in the end, was the fluidity of the race. Republican candidate Dede Scozzafava dropped out days before the election and there was a mad dash among statistical scientists to figure out who would get her voters. As PPP wrote in a post-election blog post, its findings were "way off the mark" and "very embarrassing," likely due to the "bizarre machinations in the race..."
The next worst showing among pollsters was in New York City's mayoral race, which turned out to be way closer than projected. Every single poll over the last month showed incumbent Michael Bloomberg with an advantage of at least 11 points. Quinnipiac had 50 percent of the electorate favoring Bloomberg -- which ended up being correct -- but had his challenger, Democrat Bill Thompson, at 38 percent (eight percentage points below where he ended). Marist, meanwhile, had Bloomberg with a 15-point advantage.
In the New Jersey governor's race, some pollsters ended up missing the mark. Pre-election polls showed Republican Chris Christie up by a few points. But the overall average compiled by Pollster.com had the race tied. The end result was a four-percentage point victory (49 percent to 45 percent) for Christie, pulling the rug out from under two firms that had him losing. Monmouth/Gannett and Democracy Corps, projected incumbent Jon Corzine emerging victorious by two and five points respectively.
In the Virginia gubernatorial election, the pollsters called the winner correctly, but slightly understated his margin. The Pollster.com average heading into the election had Democrat Creigh Deeds winning 41 percent of the vote to Republican Bob McDonnell's 54.7 percent. Deeds ended up right at that 41 percent mark. But McDonnell outdid expectations, gathering 58.6 percent of the vote. Survey USA hit the nail on the head when it locked McDonnell in at 58 percent and Deeds at 40 percent in its last survey on November 2.
In the end, of course, polling is an inexact science designed to give viewers a sense of where the electorate stands at a given time and not, necessarily a pitch-perfect prediction of what the result will be. And, like in every other campaign, it will be some time before one can properly assess which pollsters had a strong (or weak) cycle in 2009.
"Yes, there are several good stories about what went right and what went wrong with yesterday's polling, including some important lessons about the value of automated polling," wrote Mark Blumenthal of Pollster.com. "Some pollsters certainly did better yesterday than others. And I'm hoping to have something written and posted on that subject later today, provided that I don't get bogged down by the calls and emails from reporters wanting me to tell them, 'who was the most accurate pollster yesterday?'"