From Americans use polls to measure everything from the most popular celebrity Halloween costumes (Brangelina, 2007) to, um, our distrust of polls (68 percent of us don't believe the typical poll accurately reflects public opinion). But in an election year, polls and politics are virtually inseparable. Twenty years ago, only a handful of polls tracked presidential races, releasing new results every few weeks; today there are dozens of polls tracking the race on national and state levels.
Polling is one reason running for president is so expensive. This year, presidential candidates had already spent $28.5 million on polling and research by the end of June. What does all this money accomplish? Candidates' polls reflect trends and help them plan campaigns and raise money; media and independent polls drive coverage. Do polls shape voter behavior? Politicians and journalists think so-but how? By drawing voters to likely victors, or by rallying supporters of the underdog? By persuading voters to stay home when they think their choice is a sure winner-or a stone-cold loser? Not even pollsters know.
How can a survey of 1,000 adults accurately represent the views of 230 million? The answer is probability theory. Using a random sample, pollsters can capture a picture of the population as a whole. And the method works. Despite the closeness of the 2004 election, the average of the final major polls fell within about one percentage point of the actual outcome.
Still, polls often produce vastly different results. That's partly due to sample size and margin of error-in other words, chance. Don't obsess over one poll; watch for trends.
ANATOMY OF AN OPINION POLL
Pollsters have been telling us what we think for more than a century. Here's how the Gallup guys do it.1. Pollsters decide on the questions (e.g., Do you approve or disapprove of the way ___ is handling his job as president?).
2. Telephone numbers of potential respondents are randomly generated.
3. One thousand adults are interviewed.
4. Results are weighted to reflect nation's demographic makeup, based on U.S. Census data.
5. Editors review and interpret findings.
6. Data are published.
FLASH POINTS
The wording of questions is, along with random sampling, crucial to polling. Leading or confusing questions can produce dubious results. A poll that asked, "Does it seem possible or does it seem impossible to you that the Nazi extermination of the Jews never happened?" found that 22 percent of Americans thought it was possible the Holocaust never happened. When a clearer question was asked a year later, only 1 percent thought that.
Polls of likely voters are designed to be more accurate than polls of registered voters. They're most reliable in the late stages of a campaign as voter intentions become more fixed. "If the science of polling is sampling, selecting likely voters is the art," says Douglas Schwartz, director of the Quinnipiac University Poll. Differences in the way likely voters are identified are a common reason for variations among reputable polls.
Exit polls are in one sense the gold standard of polling because they question people who have actually voted as they leave the election site. Exit polls help explain not just who won but why. TV networks use them in projecting a winner. Critics fear that calling Eastern states based on exit polls can depress turnout elsewhere; since 1980 (when NBC declared the election over before the West Coast had finished voting), networks are more cautious in their predictions.
FORWARD THINKING
Cell phones pose a growing challenge to pollsters. Federal law prohibits the use of automated dialing-key to efficient telephone polling-to call cell phones. This means young and minority voters, who are most likely to use cell phones instead of landlines, may be under-represented. In the past, cell-only voters' political views have been similar to those of their landline-using counterparts, minimizing the problem. Will the same hold true this year? Stay tuned.Internet polls may eclipse telephone polls as cell phones, answering machines, caller ID, and resistance to phone solicitations make it harder to find willing participants. Online poll takers tend to be self-selecting (only those interested in the topic participate) and skew toward the young, the wealthy, and the better educated. But with complex weighting methods and other steps, pollsters are refining online surveys to make them more scientific and reliable.
Snail mail is poised for a comeback. The U.S. Postal Service has 99 percent of residential addresses on file-a near-perfect sampling pool. Will future pollsters combine street addresses with phones and the Internet to tap the advantages of each? "It's amazing," says Paul Lavrakas, former chief methodologist at Nielsen Media Research. "We're going back to the dark ages, to where we rely on street addresses to make sure polls are representative."
THE BACK-AND-FORTH
"There's a bonanza of polling information available to anyone with a computer [on sites like pollster.com and realclearpolitics.com]. You can be your own pollster." --Lawrence Jacobs, political science professor, University of Minnesota
"The more we rely on polls, and the more the media distort them, the more you have to worry about their impact on decision making in our democracy." --Norman J. Ornstein, resident scholar, American Enterprise Institute
"When done well and seriously, polls are a way of allowing the public to speak collectively and for their concerns to be put into the political debate." --Thomas E. Mann, senior fellow, governance studies, Brookings Institution
"A public opinion poll is no substitute for thought." --Warren Buffett, chairman, Berkshire Hathaway, Inc.
THE TIME LINE
1824 Political polling gets its start when newspapers print survey results gauging support for John Quincy Adams (who wins election decided by Congress), Andrew Jackson, and two others.1858 Debating Douglas, Lincoln cites role of "public sentiment" in politics.
1916 The Literary Digest launches first national presidential poll, mailing ballots to subscribers in five states. LD poll correctly predicts winner in every presidential election from 1916 to 1932.
1932 Emil Hurja analyzes public polling data for Franklin Roose_velt; Hurja is first polling expert to work for a presidential campaign. He later uses polls to advise FDR on building political support for New Deal.
1935 After helping his mother-in-law get elected secretary of state for Iowa, 34-year-old marketing expert George Gallup founds American Institute of Public Opinion, forerunner of the Gallup Organization.
1936 Literary Digest wrongly predicts Alf Landon will defeat FDR. LD surveyed telephone and car owners, who skew GOP. Gallup uses random sampling to forecast FDR triumph.
1948 By ending surveys too soon, pollsters miss Truman's surge to victory. Headline says it all.
1958 John F. Kennedy hires Lou Harris, first pollster to conduct private polls for any candidate for White House.
Late 1960s CBS statistician Warren Mitofsky perfects random-digit dialing technique, launching era of modern phone-based polling.
1975 Working together, CBS and the New York Times become first news organizations to conduct regular polling.
1980 NBC declares Ronald Reagan winner over Jimmy Carter three hours before California polls close-first time exit polls are used to project winner in presidential race.
1990-1991 Polls showing public support for military action against Iraq following invasion of Kuwait help George H. W. Bush win congressional backing for first Gulf War.
2000 Online polling makes its mark when Harris Interactive bests all other polls by correctly predicting tie in popular vote between George W. Bush and Al Gore.
2008 Polls leading up to New Hampshire Democratic primary give Obama a sizable lead. Clinton wins by two points, some believe with a late surge of support from women.