Newt Gingrich was not the only loser of the midterm elections. The nation's pollsters were so far off the mark, this is a good time to step up our scrutiny of an industry that has been feeding a national addiction.
Since Nov. 3, pollsters have been arguing that polls are not supposed to "predict" elections. They are "snapshots" of a given point in time, we are told. If so, the pollsters' 1998 electoral photo album must be filled with images of blurry thumbs.
The industry's credibility as a profession dates back to 1936, when George Gallup proudly claimed that a random sampling of a few hundred people could predict elections. His claim was borne out when the Gallup Poll predicted Franklin Roosevelt'sreelection, while the Literary Digest survey of 2 million readers picked Alf Landon.
It is this presumed predictive power that justifies the enormous amount of ink and air time devoted to polling results. Unfortunately, the emphasis on the horse race -- often months before election day changes the landscape itself. A CNN/USA Today/Gallup Poll shows George W. Bush beating Al Gore 57 percent to 39 percent.
Lengthy articles are written about such horse-race polls, which are then circulated by handlers and fund-raisers to convince donors and PACs that the other candidates are already out of it. This leads to more money and more endorsements, fewer resources left over for rival candidates, more positive snapshots by the pollsters, and so on, and so on. Polling thus becomes another tool in the hands of the establishment front-runners. Snapshots harden into portraits; predictions become coronations.
In a leading Minnesota poll conducted three days before the election, Hubert Humphrey III was beating Jesse Ventura 35 percent to 27 percent. But Ventura put the polling industry in a chokehold. When asked during the National Governors Association meeting what message he had for his fellow governors, Ventura replied: "They can learn from me that the American dream is still alive. They can learn from me: Don't ever believe the polls."
But the pollsters need you to believe in them. "With all due respect," said an unabashed Del Ali of the Mason-Dixon poll, "I think we were right on the money. One thing a poll is not going to predict is Hillary Clinton coming into California and the voters being as energized as they were." Mason-Dixon's final poll had showed the Boxer/Fong senate race to be a virtual dead heat; Boxer won by a 10-point margin. In polling circles, this is called "right on the money."
Writing for his peers in the "Polling Report", Humphrey Taylor, chairman of Louis Harris & Associates, confessed that, contrary to polling spin, "the possible margin for error is infinite. ... All surveys, all opinion polls ... are estimates, which may be wrong." In New York's race for U.S. Senate, John Zogby, acknowledged as one of the more accurate pollsters, showed Al D'Amato up by 3/10ths of a point on election eve. "I, personally, was kind of mesmerized by history," he said. "I saw him do it in '92 and I probably spun it more D'Amato's way than I should have."
"Spun?" But isn't "spinning" the realm of partisans and pundits, not scientists? Not exactly. Polls can be spun in myriad ways -- by changing the phrasing or order of the questions, by monkeying with the sample design, by inappropriately weighing the data.
Zogby and Taylor notwithstanding, most pollsters this year blamed not themselves but the voters for not complying with their conclusions: the turnout was too low; the turnout was too high; the unions got their voters to the polls; the Christian Coalition stayed home.
There is also the little matter of the undecided voter, who was rarely mentioned by the media but who, in many instances, became the decisive factor on Election Day. Take the poll that 10 days before the election had Wisconsin Sen. Russ Feingold losing to Rep. Mark Neumann 43 percent to 46 percent, with 10 percent undecided. The undecided voters broke Feingold's way and turned an incumbent upset into a two-point victory.
Even exit polls -- the no-brainers of the polling industry -- get it wrong. The Voter News Service reported that 22 percent of the voters for House races were from union households, compared to just 14 percent in 1994. But wait a minute. Listen to an official of the Voter News Service backtracking: "This appears to be an artifact of our change in the form of the question." In 1994, the union household question wasn't explicitly asked but merely listed as a check-off box on the back of the survey. So the bottom line is we don't know if there was an increase in union turnout. What we do know is that there was an increase in reports of an increase in union turnout.
Isn't it time we took pollsters at their word and started treating polling results as estimates? Wouldn't it be nice to see these "snapshots" posted in the back of the newspaper, next to the astrology charts?
[ Printer-friendly version ]