It's All Politics
Tue October 30, 2012
How To Read The Post-Sandy Polls
Originally published on Tue October 30, 2012 11:58 am
Hurricane Sandy's on-the-ground devastation has yet to be cataloged, and how the violent storm may affect the presidential campaign with just a week to Election Day is equally uncertain.
Will President Obama's response to the disaster help or hurt his re-election prospects? Or will the campaign's new trajectory — canceled appearances, postponed early voting — ultimately benefit Republican Mitt Romney?
Not really thinking much about that, are you?
That's why poll expert Cliff Zukin tells us that the presidential election survey world now needs to be divided: there's pre-Sandy and post-Sandy.
Any polls taken after the storm and while millions of Americans' lives remain disrupted, he says, carry with them a very real potential for accuracy problems.
"We're into a really cataclysmic event," says Zukin, former president of the American Association of Public Opinion Research. "We don't know how many people are going to be affected and for how long."
Post-Sandy surveys, says Zukin, a Rutgers University political science and public policy professor, "are going to be really suspect."
Those living in the dozen states that declared Sandy-related states of emergency will be occupied for many days — if not weeks — waiting for power to be restored, cleaning up, clearing out and dealing with insurance companies.
What's not likely to be a priority, Zukin says, is answering the phone to take questions about who they plan to vote for in the tight presidential race.
That means that surveyors will be hard-pressed to contact enough people by landline and cellphones to find a representative sample — meaning enough respondents to give pollsters the information they need to produce an accurate analysis.
And in fact, at least three national presidential tracking polls were temporarily suspended because of the storm: Gallup, Investors Business Daily/TIPP, and Public Policy Polling.
NPR contacted Zukin before Sandy hit, intending to talk with him in general about how voters should assess the blizzard of polls that have accompanied the final weeks of the presidential campaign.
While we ended up talking about the Sandy effect, Zukin also provided a few guidelines for those trying to make sense of the surveys. Here's what he advises, including his Sandy analysis.
1.) Don't compare anything pre-storm and post-storm. Anywhere from 10 to 15 percent of the country will be missing from polls for awhile as they try to get back to their regular lives. If the polls show what appears to be a pre- and post-Sandy change, it will be unclear if conditions have actually changed, or if what's changed is surveyors' ability to find people and have them respond in adequate numbers.
2.) Don't trust robocall polls that don't include cellphones. Federal law prohibits automated calling of cellphones, so pollsters who use landline robocalls should also be hand-calling cellphone users. Thirty percent of American adults use cellphones exclusively, and not sampling them will skew results in favor of people who are older, white and more conservative. For a sample of 1,000 people to be representative, about 300 of those contacted should be cellphone users.
3.) Don't compare polls done by different people or organizations. Every pollster uses different formulas to take survey results. Pollsters employ different methods of asking and ordering their questions, and use different tools (the U.S. Census, voter lists, for example) to "weight" their field results and come up with what they have determined is a representative sample. You're better off comparing over-time polls conducted by the same person or organization to detect legitimate trends.
4.) Look at the average of polls like the one maintained by Real Clear Politics. Its average of polls in both 2004 and 2008 were quite close to the final results.
But with Sandy wreaking havoc on land and in the polling world, the only remaining survey with any veracity may be the one on Nov. 6.