Pollsters Search for Reasons for Disparities in Surveys : Campaign: They explain the differences by talking about methodology. Timing and makeup of voters play key roles.

At one point this past weekend, ABC News pollster Jeff Alderman called into the assignment desk at his network and mused about the sudden narrowing in polls measuring the race for President.

“Have you ever seen the movie ‘Reefer Madness,’ ” Alderman asked rhetorically, referring to the legendary B-movie from the 1930s that warned against the hallucinatory effects of marijuana. “What you are seeing now is pollster madness.”

Since last Thursday, the polls measuring the presidential race have veered from a 19-point gap between Bill Clinton and President Bush in an NBC/Wall Street Journal poll to a 3-point spread in a Time/CNN poll.

How is this possible? Which poll should you believe? Is the race really narrowing that much?

Pollsters--and there are more of them in this campaign than ever before--explain the differences by talking about methodology. But they also say that, while the race has tightened some, it is probably still close to a 10-point gap--even before Ross Perot’s latest controversy over why he dropped out of the race in July.

And in the state-by-state polls, which are a more precise way of measuring how candidates win the presidency, Clinton’s lead still seems nearly overwhelming.

Pollsters cite several factors to account for the disparities between surveys. Timing, such as whether a poll was taken on a weekend or a weekday, can cause some differences. Some polls look at registered voters, some only at those “most likely to vote,” others at people “leaning” toward but not fully committed to a candidate.

And if readers account for the “margin of error,” the mathematical slippage built into the polls, the private White House poll that shows the spread at six points can be the same as the Gallup/CNN poll which Monday showed an 11-point gap.

To reduce confusion, polling professionals suggest averaging the polls that compare similar voter groups. The result will probably produce a margin close to double digits.

Following the second presidential debate on Oct. 15, Clinton’s lead moved up to about 15 points, according to most public polls, and some press accounts began mentioning a possible “mandate.” But after the final debate Oct. 19, “Perot really seemed to have changed a lot of people’s minds about his image,” boosting the number of Americans who thought favorably of him from 7% to more than 50%, said Andrew Kohut, who has conducted polls for Times Mirror and U.S. News & World Report.

This weekend, pontificators on the Sunday talk shows talked of the “Perot surge” in sometimes breathless terms. On “Meet the Press,” Wall Street Journal Washington bureau chief Al Hunt said he was now “much less confident than I was a week ago” about who would win the election, “because Perot is a total wild card.”

Washington Post reporter David Broder and “Meet the Press” host Timothy J. Russert declined even to offer any assessment of who might win in light of the “Perot surge.”

But the numbers were confusing. The NBC/Wall Street Journal poll surveyed voters the same days as U.S. News, Oct. 20-21. The first poll found Clinton leading by 19. The second by 6.

The next day Newsweek began polling. So did the CBS/New York Times poll. Newsweek put the race at 12 points for Clinton. CBS/New York Times put it at five points.

The differences may be due to how the questions are worded, how hard the pollster pushes someone to answer, even where the “horse race” question comes in the poll, said Ed Goeas, who polls for the Hotline newsletter.

If the survey asks people how they rate Bush as President before they ask who people will vote for, for example, Bush usually comes out lower in the poll.

In the Time/CNN poll taken late last week and published Monday, Clinton led among registered voters by seven points, 38 to 31, over Bush, with 17 for Perot. But when only those “most likely to vote” are measured, Clinton led by only three points.

That margin was so small, and so out of line with other polls, that Time issued a warning which noted “that was a significantly smaller spread than other surveys showed.”

Most of the polls have a margin of error of plus or minus 3 to 4 points, meaning that each finding could be high or low by that much. Thus when Newsweek puts Clinton at 42 and Bush at 30, Clinton could be at 38 and Bush at 34, a difference of four points not 11. Or, it could be the same 19-point spread that the NBC/Wall Street Journal poll found.

The campaigns, of course, interpret the results to suit their own ends.

In the current case, the Clinton campaign apparently sees advantage in a narrowing contest.

“The race is definitely tightening,” said Clinton strategist Paul Begala on Monday’s “Today” Show. “You know, the last thing the voters want is to be told the race is over, so let me take this opportunity to tell them this race is far from over.”

The moral was simple. If Democrats are worried, more of them are likely to vote.