Next Score View the next score

    Polls bringing more questions than answers

    Who is up and who is down in the presidential race today? There is a poll for you to consult; six, actually. What does the horse race look like in Ohio? There’s another poll to chew on. Or seven in the past week.

    Polling mania has infused media coverage of the campaign’s final days as the national obsession for clues about the outcome on Nov. 6 intensifies. And the numbers about those number-crunching firms are staggering: six daily national tracking polls; since Labor Day, about 40 other national polls by 20 outlets logged on the Real Clear Politics website; in 11 battleground states, about 240 polls by more than 50 media outlets, colleges, and commercial pollsters.

    The number of debates and disputes these polls ignite, however, is uncountable. New polls are scrutinized against trends, and media outlets such as The New York Times and Huffington Post have in-house polling analysts who develop computerized models weighted to reflect their ratings of each pollster. For the candidates’ advisers or partisan bloggers, polls are red meat.


    “Both campaigns, Republican and Democratic, have figured that polls can be used as a kind of a weapon in a discussion to keep their supporters engaged,’’ said Andrew Smith, director of the University of New Hampshire Survey Center, which polls for WMUR-TV in Manchester, N.H., and The Boston Globe. “The campaign flaks are doing more of that than in the past, criticizing polls that show their candidate doing poorly because they don’t want it to depress their supporters.”

    Get This Week in Politics in your inbox:
    A weekly recap of the top political stories from The Globe, sent right to your email.
    Thank you for signing up! Sign up for more newsletters here

    Part of the reason pollsters can easily become such political punching bags is that for all their prominence, polling is not an exact science.

    Each survey comes with caveats, including a statistical margin of sampling error of plus or minus a few percentage points and a disclaimer that the results are offered with a 95-percent confidence level. Essentially, that means one in every 20 polls could be outside the error margins — an “outlier” in the parlance of polling.

    Survey research is based on scientific principles but each pollster brings his or her own experience to the tasks of framing a sample, writing the questions, screening respondents, and weighting results in instances when it is difficult to obtain sufficient responses through random dialing of telephone numbers. Choices such as whether to measure registered or likely voters can yield different results.

    Some pollsters are more transparent in their methodology than others, and there is the ongoing debate about the accuracy of conventional live interviews versus “interactive voice response” polls, which use computer-generated calling and recorded voice prompts to produce surveys at a fraction of the cost. These polls, however, have high nonresponse rates and must tweak results to compensate for their inability to reach cellphone-only households.


    “Ironically, we have more people polling now even though it has become more difficult,” said G. Terry Madonna, who directs the Franklin and Marshall College Poll, which specializes in Pennsylvania politics and issues. One problem is declining response rates and how to compensate for them, he said.

    Engage pollsters at length about methodology and many will begin raising questions or criticizing others in the business. Not all polls are created equal, and averaging the results of good polls and not-so-good polls makes no sense, they said.

    Rasmussen Reports, the best-known of the automated pollsters, has produced the most surveys on the Real Clear Politics site. Besides a daily US poll, Rasmussen has produced multiple surveys in each of the swing states. The company also polls on consumer and cultural issues, earning revenue from subscriptions and advertising.

    What pollsters have in common is the need to have a thick skin. Madonna knows how it feels to be attacked for delivering bad polling news. Last spring, as Rick Santorum’s campaign for the Republican presidential nomination was beginning to collapse, the former Pennsylvania senator lashed out at Madonna, who has built a solid nonpartisan reputation over two decades of polling.

    Santorum called him “a Democratic hack” on national television for a poll showing the candidate had lost a huge lead to Mitt Romney before the primary in his home state. Santorum dropped out of the race the following week.


    This year, any poll or pollster straying from the pack can expect to go under the media microscope. Exhibit A is David Paleologos, director of the Suffolk University Political Research Center in Boston. He ignited a firestorm this month when he said that he would no longer poll in Florida, Virginia, or North Carolina because of “overwhelming” polling data indicating President Obama would not win those states.

    Around the time of his Oct. 9 remarks, polling in those states showed the incumbent tied, slightly ahead, or not far behind Romney in all three, but stuck at a few points below 50 percent.

    A New York magazine website blog quoted other pollsters who questioned, criticized, or ridiculed Paleologos’s remarks. One had called him “a jackass.”

    “That’s not the way I behave toward my fellow researchers,” Paleologos told the Globe. “They know better, and that amounts to nothing more than research bullying.”

    Paleologos who has headed the Suffolk operation for a decade, has polled in this cycle for WHDH/7News in Boston, stations in Florida, Nevada, and Virginia, and USA Today.

    “The great thing about this science is that at the end of the day someone is right and someone is wrong, and people have to live with their comments and their political affiliations and their opinions,” he said.

    As of Tuesday, an average of polls in the states Paleologos had written off showed Romney ahead in Florida and North Carolina and tied in Virginia.

    Brian C. Mooney can be reached at bmooney@globe.com.