Parsing polls: Nate Silver picks on target, Rasmussen not so much
The verdict is in on polling for the 2012 presidential race. New York Times statistical savant Nate Silver’s projections came very close to the actual outcome on both popular vote and the electoral college.
Gallup and Fox News favorite Rasmussen Reports — whose surveys consistently predicted a better outcome for Republican Mitt Romney — missed the mark.
Pollsters repeatedly say their surveys represent only a snapshot in time and should not be judged as predictive tools. But, inevitably, political pros and the public want to know which surveys came closest to the real-world outcome.
Silver’s FiveThirtyEight blog at nytimes.com had made a final prediction of 313 electoral votes for President Obama to 225 electoral votes for Romney. Obama had 303 electoral votes in the bag by early Wednesday morning. If the president maintains his slim lead in Florida, which stood at 47,000 votes Wednesday afternoon, he will finish with 332 electoral votes.
Silver projected a 2.5-percentage-point popular vote edge for Obama. The margin stood at 2 points, with some ballots still to be counted.
Gallup’s national tracking poll, in contrast, had shown Romney with a lead of as much as 5 points. It stopped daily tracking in the aftermath of Superstorm Sandy, and one day before the election showed the Republican nominee leading by 1 point. Rasmussen’s final tracking poll also showed Romney with a 1-point lead.
Coming closer to the actual result with their final polls were ABC/Washington Post and the Pew Research Center, both of which showed Obama with a 3-point lead two or three days before the election.
Of the big polls followed by the Real Clear Politics website, Rasmussen also had significant misses:
Ohio -- Rasmussen: dead heat. Actual: Obama by 2.
Virginia -- Rasmussen: Romney +2. Actual: Obama by 3.
Iowa -- Rasmussen: Romney +1. Actual: Obama by 6.
Wisconsin -- Rasmussen: tie. Actual: Obama by 7.
Colorado -- Rasmussen: Romney +3. Actual: Obama by 5.
Some conservative media outlets used the Rasmussen polling to prop up a narrative in the final days of the campaign that Romney had momentum and a good chance of winning the White House.
Averages of all the big polls tended to show otherwise — giving Obama a razor-thin margin in the popular vote and small but consistent leads in most of the battleground states that would decide the election.
In many of those battlegrounds, Rasmussen’s preelection surveys stood out. Other polls mostly showed Obama with a lead, sometimes larger than the polls’ margins of error.
Rasmussen has drawn scrutiny before because it employs automated surveys that do not reach mobile phones, which more and more Americans use as their primary form of communication.
The pollster also “weights” results to try to predict the party identification of the electorate. Other pollsters frown on this approach, saying one of the primary functions of the polls is to assess whether voters consider themselves Democrats, Republicans, members of some other party or no party.
Poll director Scott Rasmussen said that he thought the polls had succeeded overall by correctly projecting an extremely close race between Obama and Romney. But he also said that new techniques would be needed to better sample the public in the next presidential race.
“This race was very likely the last presidential election of the telephone polling era,” Rasmussen wrote on his website Wednesday. “While the industry did an excellent job of projecting last night’s election, entirely new techniques will need to be developed before 2016. The central issue is that phone polling worked for decades because that was how people communicated. In the 21st century, that is no longer true.”
Others pollsters came considerably closer to the actual outcome in key states.
In Ohio, a Columbus Dispatch poll had Obama up 2 points, his eventual margin of victory. In Florida, a PPP survey showed a 1-point Obama edge, also about the actual outcome.
In Wisconsin, a late October poll by Marquette University put Obama up 8 points. He won by 7. In Iowa, Obama’s 6-point win matched the edge measured by the NBC/Wall Street Journal/Marist poll in late October.
Overall, Silver’s FiveThirtyEight blog was hard to beat. The statistician/blogger might have been expected to spike the football, after taking serious abuse from some conservatives, who charged that his projections intentionally tilted in Obama’s favor.
But in a post on his site in the wee hours of Wednesday, he didn’t gloat.
“Thank you for joining FiveThirtyEight’s coverage of the election campaign,” Silver wrote. “There’s a lot more to unpack in the coming days, but first, we’re going to get some sleep and grab a beer.”
Get our Essential Politics newsletter
The latest news, analysis and insights from our politics teams from Sacramento to D.C.
You may occasionally receive promotional content from the Los Angeles Times.