Voters line up to cast their votes in Lindell School in Long Beach, N.Y. (Kathy Kmonicek / Associated…)
Which polls did best in forecasting the outcome of the 2012 election?
Most did quite well, according to several post-game analyses. But the most notable fact is that the best results came from pollsters using non-traditional methods rather than the decades-old telephone-based polls.
The three that hit the mark most precisely in their final surveys were YouGov and the Reuters/Ipsos poll, which are both Internet-based surveys, and Public Policy Polling, which uses an automated interviewing system rather than live interviewers -- “robo-polls,” as detractors sometimes call them. Those types of polls likely will be even more common in the next election cycle because telephone surveys are becoming harder and more expensive to do as more and more people stop answering phone calls from strangers.
PPP’s surveys came under sharp attack from many conservatives during the campaign because it does polling for Democratic candidates and organizations on the left side of the political spectrum, such as Daily Kos. But the firm’s results were extremely accurate. That’s particularly notable because of the number of surveys the firm performed and because its automated system cannot call cellphones, requiring the pollsters to adjust their formulas to account for them.
PHOTOS: America goes to the polls
The accuracy of the Internet polls was also striking. Many media organizations shied away from reporting the results of Internet polls because their methodology has a short track record. Unlike telephone surveys, which call people at random, Internet systems allow people to opt in to a survey population, but then match the sample for each poll to known variables such as age, race and gender and thereby mimic a random sample. The evidence suggests the process worked extremely well.
Among major media polls, all of which use traditional telephone call surveying, the NBC/Wall St. Journal, ABC/Washington Post and CBS/New York Times surveys compiled strong marks. So did surveys from the Pew Research Center.
A few polls were notable for getting the result wrong. During the course of the campaign, conservatives often cited the results of surveys by Rasmussen, whose numbers were more favorable to GOP candidates. In the end, those numbers did not hold up well. Like PPP, Rasmussen uses automated interviewing, but it also makes a guess at what the balance of party identification should be and weights its numbers to fit that model. The firm’s model appears to have assumed too large a number of Republicans.
The other firm whose numbers showed a strong skew toward the GOP was Gallup, which is probably the best-known firm in the polling business. Gallup’s daily tracking poll attracted a lot of attention during the final stages of the campaign because, until the firm suspended the poll after Hurricane Sandy, it showed Romney with a lead of as much as six percentage points.
PHOTOS: Reactions to Obama's victory
Follow Politics Now on Twitter and Facebook