Up a voter poll, without a ladder


Error: As New Hampshire showed, trying to predict the results of a state primary -- even with the best techniques available -- is risky business.

February 17, 2000|By Paul West | Paul West,SUN NATIONAL STAFF

WASHINGTON -- As George W. Bush and John McCain hurtle toward their primary showdown in South Carolina on Saturday, the latest polls are giving the Texas governor a narrow lead.

But the public might want to be wary of those numbers, because even some pollsters aren't putting much faith in the polls right now. Not after what happened in New Hampshire.

On primary night in the Granite State, while John McCain was celebrating his blowout victory, the pollsters were scraping egg off their faces.

At least a dozen organizations published polls in the days leading up to the election, including the major broadcast networks, newspapers and news weeklies. None managed to gauge the size of the McCain wave with any degree of accuracy, while several had forecast a lead for Al Gore over Bill Bradley that was larger than his slim winning margin.

One New Hampshire pollster, Dick Bennett, got it spectacularly wrong. He showed Bush with a lead over McCain right through election eve.

"I've had to eat a lot of crow in the past," says Bennett. "I know how to make it taste pretty good."

A final pre-election poll by CBS News showed McCain ahead by a scant 4 percentage points; he won by 18. On the Democratic side, CBS showed Gore 16 points up on Bradley; he won by just 4 points.

CBS wasn't alone in misjudging the electorate. Five polling groups tracked daily voter sentiment leading up to the election. None of their final polls gave McCain a lead greater than 12 percentage points. The average McCain edge was 7 points, less than half his actual margin of victory.

As the election returns were rolling in that night, a friend asked Dotty Lynch, CBS News' political editor, why all the polls had been so far off base. She came back with a stock answer that pollsters often use in such situations.

"A lot of people made up their minds on Election Day," she said with a hearty laugh.

There was at least a grain of truth in that reply. Much of Bradley's support came from voters who had made up their minds as they headed for the ballot booth.

Still, the pollsters' bad election night points up some of the difficulties of polling in presidential primaries. In a post-New Hampshire statement, the National Council on Public Polls, which seeks to raise the standards of the profession, describes those problems as "almost unsurmountable."

Among the difficulties: Turnout in primaries is often low, making it tough to identify those who will actually vote. Also, the campaigns are shorter than general election races, and many voters can be swayed to almost any candidate.

"Most pollsters understand these pitfalls, and they do polls anyway," the NCPP statement concludes. "Our advice: caveat emptor [let the buyer beware]!"

Other factors might have also contributed to the skewed poll results. They include:

Polls are only a snapshot of sentiment at the time they are taken, not a predictor of results.

The poll that came closest to matching the actual victory margin in the Republican race was conducted two weeks before Election Day. But that might have been largely a matter of luck, since there were wide swings in voter attitudes in the days that followed.

"I didn't believe my own numbers," says Whit Ayers, a Republican pollster whose survey showed McCain with a 17-point edge.

In general, the closer a poll is taken to Election Day, the more accurate it is likely to be. Still, things change in campaigns.

And people change their minds, often when they step into the voting booth. This is especially true in a state such as New Hampshire, which is known for its volatile and unpredictable electorate.

There is an element of educated guesswork to the "science" of polling, though some pollsters don't like to acknowledge it. That shortcoming was on display in New Hampshire, and it could be again this weekend when the votes are counted in South Carolina.

The problem boils down to turnout, the factor that ultimately determines who wins. Pollsters use devices to try to figure out who will vote, including asking a series of questions designed to screen out nonvoters.

But pollsters can never be certain which respondents will vote. Some of those who say they will probably vote never do.

The "X" factor in New Hampshire, opinion analysts realized from the start, would be how many independent voters -- who outnumber either Republicans or Democrats in that state -- would actually cast ballots.

Complicating that equation, independents could vote in either the Republican or Democratic primary. And they didn't have to decide that until Election Day.

"In a state where independents can participate in party primaries, it's anybody's guess as to how many of them are going to vote," says Warren Mitofsky, a polling consultant for CBS and CNN.

Bennett, the New Hampshire pollster who missed the McCain wave, guessed wrong on independent turnout. Based on his 23 years of experience in the state, he factored in a figure that turned out to be barely half of the huge independent vote, which went heavily for McCain.

Baltimore Sun Articles
Please note the green-lined linked article text has been applied commercially without any involvement from our newsroom editors, reporters or any other editorial staff.