Total Pageviews

Thursday, November 1, 2012

Examining the Effect of the Storm on Polling

In the final days of this intensely close election season, polling has taken on an added layer of uncertainty with millions of people stranded without electricity and without phones in several states. It may seem even silly or ridiculous to those experiencing the worst storm damage in their lives to talk about polling (or for that matter receive a survey call on a cellphone whose charged battery is indeed a lifeline), but for the data-obsessed and for pollsters, storm Sandy has complicated matters just a bit.

Some pollsters have been forced to scrap plans to poll certain local races or Congressional districts even if the contests are highly competitive, because a sizable portion of the electorate may not have electricity or is dealing with severe weather damage. Others, like The New York Times and CBS News, whose pollsters conduct interviews in-house in Manhattan, on Sunday cut polling off by a few days ahead of the storm and released a national survey with a smaller sample than expected. The Times probably will not go back in the field at all before Election Day. (Separate polls of Ohio, Virginia and Florida conducted in collaboration with Quinnipiac University and CBS News were also stopped on Sunday, but they had been under way since Oct. 23.)

Fox News, as another example, noted in releasing its national poll on Wednesday that “interviews in areas affected by Hurricane Sandy were completed before Monday evening, when the hurricane made landfall.” The Marist College Institute for Public Opinion wrapped up polling in New Hampshire â€" one of the battleground states where it has been regularly conducting surveys with NBC News and The Wall Street Journal - on Monday night, before any power failures occurred, according to Lee M. Miringoff, director of the institute.

Other major organizations are making adjustments as they gear up for the finish line on Tuesday. Gallup, for example, is resuming its daily tracking poll on Th ursday, after deciding Monday to stop polling when it determined that “the storm compromised our ability to provide a nationally representative sample,” said Frank Newport, editor-in-chief of Gallup. And even though it is the last week of pre-election polling, Gallup has decided not to issue reports day to day, because of the three-day hiatus, “or lacuna,” as Mr. Newport labeled it, while the surveys were suspended. Adding fresh data to last week's seven-day rolling tracker “didn't seem to make sense or to provide valuable information to the public,” he added. So Gallup plans, at least for now, to release Thursday-Sunday tracking results this weekend.

A few polling organizations have stayed in the field throughout the storm and its aftermath, and some continued daily tracking polls. One set of polls that meets The Times's polling standards, those by ABC News and The Washington Post, has released a survey each day this week at 5 p.m., reporting a four-day wa ve of polling of 1,288 likely voters. (Frankly, the tracker has moved little in the last few days, what with the presidential race extremely tight.)

Cliff Zukin, a former president of the American Association of Public Opinion Research (Aapor), cautioned against interpreting too much from tracking surveys conducted after the storm and also against making strong comparisons to polls taken before Hurricane Sandy made landfall Monday evening. Mr. Zukin, a professor at Rutgers University, said that for some, the storm had achieved a prominence in terms of public interest that is as dominant if not more so as the election.

Of the tracking polls, he said, “We should be less confident in them in that there are two major things going on in the country right now rather than just one major thing. So polls may pick up a change in vote intention because citizens believe Obama's doing a good job with the storm or it may be that we have m ore response problems than we had before.” The storm and its aftermath, Mr. Zukin added, is “just a big, big high-profile item that is out there that a week ago was unthinkable.” He said the storm could not only affect people's frames of reference but also, in the end, their willingness to actually vote.

Mr. Zukin said he was less concerned about surveys that span the next several days, because adjustments can be made through various callback procedures and other means.

For those surveys conducted now through the rest of the weekend, some analysts and pollsters have added even more layers of data sets to their analysis, to take into account the possibility of extremely poor response rates or high refusal rates in those areas like New Jersey, where two million people are still without power, many without landlines or working cellphones. Power grid reports and storm damage reports, overlaid by ZIP codes or county, have been pored over to try to ensure that th e surveys collect a representative pool of voters.

Pew Research Center went into the field on Wednesday night and hopes to complete its final pre-election survey by the end of the weekend. Scott Keeter, the director of survey research at Pew and a former president of Aapor, said that Pew's polling team had been analyzing what portion of the public would be unreachable because of power problems and planned to monitor the results and assess the response rates from affected counties each night.

Whether a national survey conducted now would show anomalies by missing residents in hardest-hit New Jersey, which in 2008 went comfortably for Barack Obama, or other areas without electricity or absent those who left evacuation zones, is an unknown and will have to be watched closely, Mr. Keeter said. Given the electorate in New Jersey, it is possible that any omission of Democrats â€" there or in other states â€" because of the storm could be offset a bit by gaps in reachi ng people without electricity or facing storm damage in places that are more Republican, like in the mountains or in eastern Ohio, he said.

“We debated delaying a day, we debated canceling the poll if we thought it too biased,” Mr. Keeter said, adding that the latter would still be an option. Perhaps more likely, after reviewing the composition of this survey's sample in comparison to previous Pew surveys, would be to determine whether any type of correction or weighting adjustment would be worth doing to achieve a representative sample.

“Is there some correction that could be applied that would do more good than harm?” Mr. Keeter asked. “That's always a question when you start adding funky weighting or things that you haven't done before.” But, like many other major independent pollsters, Mr. Keeter said Pew would not weight the results to party identification from previous election turnouts.

Other pollsters have had to switch late-election pla ns altogether, and to cancel scheduled surveys in a few hotly contested races. Donald Levy, director of the Siena Research Institute, outside Albany, said he decided to forgo polling in the 1st Congressional District on Long Island, where Representative Tim Bishop, the Democratic incumbent, and Randy Altschuler, a Republican businessman, are locked in a rematch that Mr. Altschuler narrowly lost two years ago. Siena has also canceled polling in the State Senate District 15 in Queens. “Obviously it threw us into turmoil,” Mr. Levy said, and so they shifted some resources into races upstate where voters were largely unaffected by the storms. He was still assessing conditions in Westchester County, where the 18th Congressional District is situated.

“Never in my career have I spent time studying power outage reports,” Mr. Levy said. “But here I am collating the power outage reports across New York State to try to determine if we can surgically go in, switch our pl ans and surgically go into some districts.”

Lastly, some of the pollsters interviewed (several of whom were without electricity themselves this week) wondered aloud how those in storm-affected areas would feel about actually voting next week, if they remain isolated without electricity or Internet service, and if their polling places are shifted around or consolidated â€" conditions that could cause even greater confusion.

At Edison Research, which conducts next week's exit polls for the National Election Pool that includes the major television networks and The Associated Press, some telephone polling is already under way nationally and in states with significant percentages of early voting patterns to augment actually exit polling at precincts nationwide on Tuesday. To a great extent, even though Edison was in the field before and after the storm, and in states that were not affected Monday and Tuesday, there was little disruption of interviewing, said Joe Len ski, executive vice president of Edison.

On election night, however, Mr. Lenski said, “I will not be surprised if turnouts are way down in New Jersey and New York and other places.”

Kate Phillips is the editor of News Surveys and Election Analysis for The New York Times.