While Nate Silver is reaping accolades for his pitch-perfect predictions of the electoral vote in every state, his model analyzed polls conducted by many competing polling firms.
So which polling outfits were accurate?
The most prominent polling firm is Gallup, whose daily tracking poll received much press when it showed Romney with as much as a 7-point lead in the popular vote in the waning days of the election.
It's obvious now that those numbers weren't accurate. The Gallup tracking polls aggregates data taken over the previous seven days, which should tend to smooth out spikes, but it can also amplify errors in methodology, and that seems to be the case here.
Rasmussen was also off-base, getting six out of nine swing state predictions wrong. I've talked before about the issues inherent in their "robo-calling" methodology. By law, their automated system can't contact cell phone users like more expensive live-calling operations can, so they only reach landline users, and their automated recordings get lower response rates.
Younger, and therefore more liberal, voters tend to only have cell phones, and often won't stay on the line for a recorded message, so Rasmussen polls skew slightly conservative, about 2 points according to Nate Silver, whom I think we can all agree to trust now.
Rasmussen's national daily tracking poll also uses a rolling average, though only three-day one. It was one of the few polls showing a Romney lead even up until the election.
Most notably, I think, is that Rasmussen is the only firm that showed anything other than an Obama-lead in Ohio during the last three weeks of the campaign. I've mentioned that it was possible everyone else was wrong and Rasmussen was right.
They weren't.
Other polling outfits fared much better. The NBC/Wall Street Journal poll got seven out of nine swing states calls correct.
The CBS/New York Times/Quinnipiac poll correctly predicted four out of the five swing states they covered.
As for the popular vote, we won't have the final totals for perhaps several weeks as absentee ballots trickle in, but it looks like Obama has a lead between 1.5 and 4 points.
The Pew Research Center and the ABC News/Washington Post poll both predicted a 3-point Obama win, and NBC/WSJ gave Obama a 1-point margin of victory.
Most election coverage sites, like RealClearPolitics or FiveThirtyEight, aggregate all the polls, increasing the sample size and reducing the margin of error. And that approach really seems to have worked.
Whatever the shortcomings of individual polls, taken as a whole they were spot-on. Next election, ignore the pundits and trust the polls.
- Contribute to this Story:
- Send us a tip
- Send us a photo or video
- Suggest a correction
