Political polling is supposed to be scientific. But a lot of people in Florida have their doubts. Just look at the United States Senate race. According to recent polls:
Sen. Paula Hawkins, the Republican candidate, has just staged a remarkable comeback and has pulled eight points ahead in her race with Democratic Gov. Bob Graham.
Or, according to other polls, Governor Graham has blocked Senator Hawkins's political offensive and still leads by a comfortable 10 to 13 points.
Or, some polls indicate, Graham is trouncing Hawkins, who could lose this race by as much as 20 percentage points.
Voters are perplexed. One day Graham is reported ahead by 12 or 18 points, while another poll shows the margin to be only three points.
Pollsters are defensive. They deny there's been any shoddy work, or any funny business. But politicians are accusing each other of covering up their true poll numbers.
Political polls are crucial to the campaign in Florida, as elsewhere, because of their impact on fund raising. The big-money boys don't want to waste their cash on Hawkins or Graham if either is a sure loser. Political funds flow toward likely winners.
That's one reason Graham's campaign cried ``foul'' when Hawkins's pollster, Dick Morris, released a poll two months ago. The poll showed that Graham led Hawkins by less than 1 percent.
``Nonsense,'' harrumphed a Graham aide. Other polls were showing Graham 10 to 20 points ahead.
Yet the confusion hasn't ended. Morris's poll was followed by one from the respected Market Opinion Research, headed by GOP pollster Robert Teeter. That showed Graham leading by only three points, 48 to 45. Then came a Tampa Tribune poll showing Graham ahead by only two. And recently, a Palm Beach Post poll showed a spread of only four.
On Friday, the whole race appeared to be turned suddenly upside down as Hawkins leaped ahead of Graham 48 to 40 in a poll by USA Today.
Meanwhile, a poll conducted a short time before for the New York Times Newspapers in Florida showed Graham still 13 points ahead.
What's going on here? Are some polls inaccurate? Do pollsters, or politicians, lie?
Florida's polling conundrum is nothing new. In the 1984 presidential race, the polls all indicated that President Reagan was ahead, but the percentages varied widely.
Gallup, however, got it right that year. George Gallup Jr., head of the highly respected Gallup Organization, says shoddy polls are rare, and he doubts there is any intentional distortion in this year's Florida Senate race, or in 1984 presidential polling.
He suggests one probable reason for the great differences is technical. The polls are measuring different things. One poll might measure the views of registered voters, while another measures likely voters. That small difference can make a large difference in results.
Small differences in the wording of questions can also change the results of a poll. For example, does the questioner identify Graham as a Democrat, and Hawkins as the incumbent Republican senator? To do so could shift 2 or 3 percent of the voters.
The accuracy of polls can also depend on where they are conducted. Keith Frederick, who works with William Hamilton & Staff, Graham's pollster, calls Florida ``one of the toughest states in the country'' to get good polling results.
Mr. Frederick says so many people move in and out of the state, live there part-time, or have just moved there that it's extremely difficult to figure out who will vote.
Furthermore, voting behavior is becoming difficult to predict as the state grows, because it now consists of five ``different states,'' Frederick says. Pollsters must figure out how much weight to give each of those major regions.
If there's an unexpectedly large voter turnout in Hispanic-dominated south Florida, for example, it could completely undo the results of even the most carefully crafted poll.
G. Donald Ferree, associate director of the Roper Center, suggests that part of this year's problem in Florida may be the great popularity of both candidates. ``A lot of people may honestly feel torn,'' Mr. Ferree says.
He compares it to going to a restaurant and liking two dishes. Ferree says, ``It's not that you don't care; you really are torn.''
Graham and Hawkins offer that kind of choice. Each has a popularity rating between 70 and 80 percent. Voters apparently would be happy to have either of them. So on one day, they may say ``Graham,'' when asked. On another, ``Hawkins.'' It can make accurate polling very difficult, Ferree suggests.
All this can be discouraging to pollsters, such as Harry O'Neil, vice-chairman of the National Council of Public Polls, a watchdog group. He says:
``Obviously we are always concerned professionally when polls are jumping all around. If polls are done the same way, if the sampling is done essentially the same way, if pollsters are screening those they interview similarly, . . . you ought to have results that are fairly close.
``To the extent that somebody is cutting corners, not doing enough interviews, maybe not doing any callbacks, not screening anybody out at all, and doing it on the cheap, then you leave yourself open for this kind of fluctuation.''
A good poll, sampling 700 people, with an accuracy of 95 percent, should cost $15,000 to $20,000, Mr. O'Neil says.
As for the befuddled voters of Florida, there seems little hope that this confusion over polls will end soon. So they'll just have to wait for that one, final poll with absolute accuracy: election day, Nov. 4. Next: Women in politics