Engineering Inc. - March/April 2008 - 11

The Politics of Polling Political pollsters outline the mechanics of polling and reveal keys to understanding presidential polling By Maureen Conley P olling plays such a prominent role in elections that recent failures to accurately gauge Hillary Clinton’s support in New Hampshire, or Barack Obama’s in South Carolina, raise the question: What are these pre-election polls really worth? Frank Newport, editor-in-chief at the Gallup Poll, says polling is “a statistical technique to estimate what we would find if we interviewed everybody in a population,” providing a snapshot of the public’s preference at the time of the poll. The closer to the election, the more accurately the poll can predict the outcome. April Radocchio, assistant director of the Quinnipiac University Poll, says the best polls follow standards set by the American Association for Public Opinion Research (AAPOR). To conduct a valid probability sampling poll, AAPOR’s website says, the pollster first Frank Newport must define the population to be sampled: U.S. citizens, likely or registered voters, state residents, Republicans, Democrats and independents are among the most common categories, and each can be broken into subgroups by age, gender, ethnicity and so on. The pollster also must select a “representative” sample—meaning one that represents accurately the population at large. Everyone in the sample has to have a chance of being selected for the poll. They do not, however, need to have an equal chance, which is where margin of error comes in. Newport says random samples can be compared demographically to large population characteristics, such as those maintained by the Census Bureau, and adjusted if necessary. Margin of error is “the price you pay for not talking to everyone in your population group,” AAPOR explains, and depends mainly on sample size. What’s more, Newport says, the margin does not change dramatically once samples reach 1,000. Gallup says it sometimes uses 2,000person samples so that subgroup answers can be reported more accurately. What Went Wrong? So who’s to blame for the woefully inaccurate polls prior to primaries in New Hampshire and South Carolina? Pollsters have been conducting postmortem examinations of New Hampshire data, reinterviewing poll participants to better understand what happened. New- Scott Keeter port says those studies will take time. A special AAPOR committee says it plans to report its own findings in April 2008. Theories abound. Some point to the large number of voters who made up or changed their minds on election day, or extraordinarily high voter turnout. Others suggest the compressed primary schedule did not allow polls sufficient time to reflect quickly moving events on the ground. Still others point to the Bradley effect, referring to the 1982 California gubernatorial campaign of Los Angeles’ AfricanAmerican Mayor Tom Bradley, who narrowly lost after leading in pre-election polls—giving rise to the theory that some white voters might not admit they oppose a black candidate. Or perhaps Clinton’s support was not fully captured because her supporters were less likely to participate in polls. Scott Keeter, director of survey research at the Pew Research Center, says that, for the most part, polls tend to reliably predict elections. But he says they are “best used for understanding why people voted the way they did. Was it a vote for continuity or change? Was it based on certain issues as opposed to others?” asks Keeter. Fritz Wenzel, director of communications at Zogby International, warns polls are not a good predictive tool and “should not necessarily influence what people think” or how they vote. Gallup employs statisticians and methodologists who constantly work to make its polls as accurate as possible. Zogby performs quality spotchecks by calling 2 percent of a sample a second time to verify that answers are accurately tallied. Pew tries to call each number multiple times, on different days and at different times, to give everyone in a sample a chance to be polled. Interviewers are well-trained and regularly monitored, and Pew pretests its surveys before they are deployed, says Keeter. Keys to Victory Asked what demographics are important in this presidential campaign, not all pollsters agree. In a close election, Keeter says, it could be women, and how much they rally around Clinton if she is the Democratic nominee; independents, and whether their skepticism about Clinton can be overcome; evangelical Christians, who gave 78 percent of their votes to Bush in 2004, but might not support John McCain in the same measure; and the black vote, and whether a highly energized black electorate could offset resistance to Obama elsewhere. Newport believes swing voters will again hold the key, since partisans are likely to vote their ticket. n maRCh / apRIl 2008 ENGINEERING INC. 11

Engineering Inc. - March/April 2008

Table of Contents for the Digital Edition of Engineering Inc. - March/April 2008