Photo: Suffolk University
One month before Election Day, David Paleologos threw himself into the poll-heavy discussion of the 2012 campaign with a bold prediction.Paleologos, the director of the Suffolk University Political Research centre, said that he would no longer poll three key swing states — North Carolina, Virginia and Florida — because he was certain Republican Mitt Romney would win them.
That prediction, obviously, turned out to be horribly wrong. Romney won North Carolina, but Romney lost Virginia by nearly 4 points — and one of the biggest surprises of the election came when President Barack Obama won Florida.
In an interview this week, Paleologos said he didn’t intend for the proclamation to come across as so audacious. In choosing not to enter the field and poll the states, he said he was simply following a rule to which many pollsters subscribe: the incumbent rule.
When Paleologos appeared on Fox News in October and made his prediction, Obama was leading or tied with Romney in both Virginia and Florida. But he was stuck at 47.8 per cent and 47.3 per cent in the states. Historical precedent generally suggests that an incumbent still lagging below 50 per cent at that point in the campaign won’t win — hence, the “incumbent rule.”
“The incumbent rule didn’t stand,” Paleologos said. “It was disproven and tipped on its side.”
For other races, it held up. In early polling for the Massachusetts Senate race, Paleologos thought that Scott Brown would be in trouble of losing — not because of his head-to-head numbers vs. eventual winner Elizabeth Warren, but because of his standing against potential Democratic candidate Marisa DeFranco. Brown led DeFranco by a 49-28 margin in Suffolk’s first gauge of the race — below the key 50 per cent mark.
“This election rewrote the case law,” he said of the Obama-Romney race in Virginia and Florida.
How did this happen? Paleologos suggests a few reasons.
The first is that some common poll assumptions and voter screens were off. For instance, Gallup’s much-maligned polls used a likely voter screen that missed badly on its prediction of the electorate.
It goes hand in hand with a poll Paleologos and Suffolk conducted in August, in conjunction with USA Today. It found that unlikely voters could “turn a too-close-to-call race into a landslide” for Obama. The voters disproportionately favoured Obama by more than a 2-to-1 margin, and they were more likely to come from the young, African-American and Latino demographics that rode him to victory.
“The Obama campaign brought out the casual votes,” Paleologos said. “African-American, young, Hispanic, single women — they were probably not picked up by a lot of likely voter screens and the Obama campaign found a treasure trove of them and turned them out.”
Other pollsters, however, don’t subscribe to the incumbency rule. Tom Jensen, the director of Public Policy Polling, a firm that correctly predicted all 50 states, said that it has been outdated since the 2004 presidential race between incumbent George W. Bush and John Kerry.
“I don’t think the incumbent rule is really valid,” Jensen wrote in an email. “Democrats sort of used that as their crutch in 2004 and now a lot of Republicans did in 2012 and in neither case was it validated.
“I don’t know if it was ever true but it definitely isn’t now. I think pollsters generally shouldn’t make any grand pronouncements that they don’t have the data to back up, so if your polls say Obama’s leading in a place it’s an error to go out there and say Romney’s going to win in it. I think we’re usually better off sticking to our numbers and not trying to play pundit.”
For his part, Paleologos said that Suffolk would re-evaluate its likely voter screen and the way it approaches a decision on whether to poll certain states.
He might consider adopting a “looser” likely voter screen. In Suffolk’s case, that means that if a respondent says they are 50-50 on possibly voting, the researcher would continue the call — not hang up.
“I think a lot of these polls like Gallup found people who were going to vote, but they just didn’t indicate it, or indicate it at the time,” he said. “I mean, Gallup registered-voter surveys came closer than likely voter outcomes. The tight screen didn’t work well.”
NOW WATCH: Briefing videos
Business Insider Emails & Alerts
Site highlights each day to your inbox.