Exit the Polls?

Since 1936, when George Gallup first applied scientific polling methods to an American presidential election, opinion surveys have been a prominent part of our political life. Of course, there have been doubters from the start. In the 1930s some critics in Congress wanted to ban the polls, fearing they'd create "bandwagon" effects and distort electoral outcomes. Sometimes the critics were vindicated, as in 1948 when Dr. Gallup followed the rest of the experts in wrongly predicting a Truman defeat.

That missed call prompted a Gallup-led review of polling industry practices.

Polls have persisted despite setbacks and critics, becoming an ever larger factor in media coverage of elections. And they're not likely to fade from the scene now, even after a 1996 performance that again raises doubts about how polls are conducted and used.

For starters, this was a year of multiple "exit poll" backfires. Close, highly important senatorial races, like Democrat Dick Swett versus incumbent Republican Bob Smith in New Hampshire, were badly called on the basis of this instant opinion sampling of people who have just cast their votes. On the basis of early calls in this and other races, one might never have foreseen the GOP's eventual success in maintaining control of Congress. Better for their credibility had the networks simply waited, traced actual returns, and called off their high-tech race to be first to declare a winner, even if wrong.

Perhaps more disturbing, major polls misfired on the presidential race itself, giving President Clinton a much wider margin of victory over Bob Dole than the 8 points by which he finally won. This reignites controversy over the effect of polls on voting behavior. Would the contest have been even closer had the polls not been asserting, right to the end, that there was virtually no contest? Would the national turnout, the lowest since 1924, have been greater?

A lot of factors could figure into the polls' sliding accuracy (they weren't very good at predicting the Bush landslide in 1988 either). But a prime suspect is the inundation of the public by surveys. Polling scholar Everett Carll Ladd has pointed out that from Sept. 1 through Nov. 4 of this year national polls plied voters with queries about their presidential preferences 300 times - twice as often as the same period in 1992. Meanwhile, people are being flooded throughout the year with telephone surveys about their taste in toothpaste, life insurance, magazines, whatever. The rising refusal rate faced by pollsters reflects a disgust with polls, and it makes the pollster's task of getting an accurate sample more difficult.

It's time, clearly, for another look inward by pollsters. Some serious soul-searching will be needed to reburnish George Gallup's article of faith - that public opinion polling, wisely used, should enhance democracy, not harm it.