Campaign ’96 / MEDIA : Journalists Wield Polling Sword at Their Peril : An invaluable tool for divining the vote proved unreliable in Iowa, controversial in New Hampshire.

Sometimes polls help journalists figure out the political landscape. Sometimes they lead them astray. In the two major votes so far this campaign year, the polls were off the mark in one race--Iowa--and caused controversy about early network projections in the other--New Hampshire.

“Entrance polls” taken as voters went to cast their ballots at Iowa’s Feb. 12 caucuses led to early declarations by some news organizations that Senate Majority Leader Bob Dole would win big. When Dole only squeaked ahead of commentator Patrick J. Buchanan by 3 percentage points, the news community was chastened. This week in New Hampshire, there were vows of caution about the polls, especially when an avalanche of data in the last few days showed the GOP primary too close to call.

NBC’s Lisa Myers, for example, balked when she was asked over the weekend what her gut feeling was about who was winning. “My gut tells me to keep my mouth shut,” she said.

CNN’s Bruce Morton saw the pre-primary polls as a signal to wait for the vote. “When it’s this close, none of our traditional omens and symbols help at all,” Morton said shortly before the vote started coming in Tuesday. “I have no feel at all for how this one’s going. There’s a lot of ‘gee whiz’ going on around here.”

But on primary night, the networks could not resist the race to be first to name a winner. With less than a quarter of the vote in, CNN declared at 5:23 p.m. PST--less than 30 minutes after the polls had closed--that Buchanan had won. Dole would come in second, they said, and former Tennessee Gov. Lamar Alexander would be a respectable third. The other networks followed soon with the same predictions.

As it turned out, the networks were right. Buchanan won narrowly, with 28% to Dole’s 27% and Alexander’s 23%. But for some polling experts watching the same data, the network calls were dangerously early.

After Buchanan was declared the winner, the race got closer, with Dole a scant 600 votes behind the conservative commentator at one point.

Significantly, Murray Edelman, executive director of the Voter News Service that does the exit polling for the networks, Associated Press and other clients, delayed declaring the race himself until about 7 p.m. PST.

“The estimates I saw on my screen suggested that the race was a lot closer” than the early vote showed, Edelman said. “My view was that the networks had jumped the call.”

Some network experts, who analyze the exit poll numbers themselves, said their readings of the data showed that Edelman was being too conservative. But one network political expert said Wednesday that the early projections were a big gamble “and we were just lucky.”

Even in the year’s first election--the Feb. 6 Louisiana caucuses in which only Buchanan and Texas Sen. Phil Gramm competed--the entrance polling of voters fell short. Although Buchanan’s victory was correctly predicted, the polling exaggerated his margin.

Polling has always been a controversial business within the journalism community, where many argue that the polls often “drive” the news and promote a kind of horse-race-style journalism. But the number of polls has continued to grow, with many news organizations hiring their own pollsters for the big contests. In New Hampshire alone, there were about nine polls being taken almost constantly over the last few months.

“There were a lot more of these polls this time than ever before,” said Larry Sabato, a professor of government at the University of Virginia who has criticized the media for depending too heavily on such surveys. “But the nice thing was that they tended this time to cancel each other out.”

A check of the final poll results compared to the actual vote shows a variety of predictions--some with Dole as the winner, others with Alexander lagging too far behind. Indeed, most veteran journalists who stared at the array of bouncing numbers decided that these surveys were sending one clear message: a large yellow caution light.

Several tracking polls--telephone surveys that “track” voters over several days--were at odds in the days before New Hampshire’s vote. Election eve polls for the Boston Globe and WBZ-TV in Boston, along with one for USA Today and CNN, came closest to getting the final results right.

ABC-TV’s last tracking polls showed Buchanan beating Dole by a hair, but Alexander a distant third. Others being reported to the media--including the American Research Group, which serves corporate clients in New Hampshire, and American Viewpoint, which did some polling for Gramm’s failed candidacy--showed Dole in the lead by 6 to 7 points. And the Concord Monitor’s last poll ending on Saturday showed Dole first, Alexander second and Buchanan third in a close race.

For many critics, however, the problem with the polls is that they keep the journalists focused on the contest--especially when the surveys are funded by media organizations that then feel obligated to use and promote them.

Others disagree that the media should sidestep the horse race. “That’s bunk,” said ABC political director Hal Bruno. “Telling people who’s winning and losing is what journalists are supposed to do. We’re not academics; we’re not the scholars. We’re journalists.”