What do numbers mean? Answer lost in poll 'orgy'

Go ahead, pick a number.

Bill Clinton leads by 12 points or 11 points or 10 points or 8 points or 7 points, although it may be as many as 19 or as few as 2.


Who knows for sure? Not the poll-takers, and certainly not the voters. Less than a week before Election Day, the nation is drowning in a sea of numbers.

"This is the worst orgy of polling in American history," says Larry Sabato, a political analyst at the University of Virginia who has written extensively about media behavior. "There have never been so many media polls, taken so frequently. It's a travesty. Journalists indulge in worthless characterizations based on polls that may not be accurate."


Even the pollsters are crying "caveat emptor."

"Voters get irritated, with good reason," says survey expert Albert Cantril, a trustee of the National Council on Public Polls. "It's saturation bombing. The rule of thumb should be: Don't take any poll as the God-given truth."

Pick a number, because anyone can play: NBC, ABC, CBS, Cable News Network, Time, Newsweek, the Wall Street Journal, the New York Times, the Washington Post, Times Mirror.

And don't forget that 3-point margin of error: Mr. Clinton's lead in the NBC-Wall Street Journal poll could be as low as 16 or as high as 22. His lead in the CBS-New York Times survey could be as low as 2 or as high as 8.

"An accurate poll sample is akin to having a blood sample taken from your body," says Democratic pollster Alan Secrest. "The blood in that vein is supposed to be like the blood everywhere in your body. That's how it's supposed to be with polls. But some can be egregiously off the mark."

One problem is the growing popularity of overnight "tracking" polls that may be less reliable than surveys conducted over two or three days. Tracking polls get fresh numbers each day, and those numbers are averaged with those from the most recent days.

Yesterday, the CNN-USA Today tracking poll of 1,217 likely voters showed Mr. Clinton at 40 percent, Mr. Bush at 38 percent and Mr. Perot at 16 percent. With a margin of error of 3 percentage points, that added up to a dead heat.

But ABC said yesterday that its tracking survey of 900 likely voters had Mr. Clinton ahead 42 percent to 35 percent for Mr. Bush and 20 percent for Mr. Perot. Its margin of error was 4 percentage points.


And an NBC News-Wall Street Journal one-day poll of 576 voters was similar, though it is not a tracking poll. It put the race yesterday at 43 percent for Mr. Clinton, 36 percent for Mr. Bush and 15 percent for Mr. Perot. It had a margin of error of 5 percentage points.

"This is overkill on the part of the media," Mr. Cantril says. "The average person tends not to distinguish between tracking polls and the better polls. And there's a big difference."

By definition, there's not much time to conduct a tracking poll, which means there's no time to call back people who didn't answer the phone the first time around. Mr. Cantril says it is important to reach those people to ensure a representative sample, since people under 30, particularly men, are more likely than others to be out at night.

Another problem is that the polls tend to undercount the undecided. David Moore, a New Hampshire pollster, says that indecisive voters are all too often pressed to declare a preference when they don't have one.

In most of the new surveys, fewer than 10 percent of the voters are listed as undecided. John Benson, a senior opinion analyst for the Roper organization, doesn't believe that. "There are a heck of a lot more people that are undecided," he says.

Then there's the fact that a growing percentage of people simply refuse to speak with pollsters. The latest "refusal rate" exceeds 30 per cent, and it's particularly strong among racial minorities, low-income people and the elderly.


The best surveys will correct for that underrepresentation, Mr. Benson says, "but if you're doing an overnight poll and you've got limited resources, it's easier to just dial another number."

So, in light of all this, should voters discount the fact that Mr. Clinton appears to be ahead with six days remaining? Probably not. In fact, Mr. Benson and others suspect that his lead among "likely voters" might be greater than those polls indicate.

When pollsters screen for likely voters, they usually omit those who haven't cast a ballot in recent elections. But Curtis Gans, who directs the Committee for the Study of the American Electorate, says such a criterion "doesn't tell you about the surge of interest" in this election.

But nobody really knows. Republicans are praying for a repeat of the British election in April, when Conservative John Major was elected prime minister although all the polls predicted his defeat.

It has happened here before.

After President Harry S. Truman stunned "front-runner" Thomas E. Dewey in 1948, essayist E. B. White declared, "We are proud of America for clouding up the crystal ball."