In an article yesterday on polling in the New Hampshire primary, it was stated incorrectly that The Washington Post-ABC News poll never reported a lead of more than one percentage point for Sen. Robert J. Dole (R-Kan.). In fact, Tuesday's Post reported that the average of the previous three days of polling gave Dole a 3-point lead, although a survey of 405 voters Monday night showed the race even. Vice President Bush won the voting by 9 points. Yesterday's article also described as a CBS-New York Times poll a survey conducted only by CBS. (Published 2/ 19/88)

For Vice President Bush and his supporters, Tuesday's 9-percentage-point victory over Sen. Robert J. Dole (R-Kan.) in New Hampshire was a delightful surprise; for Andrew Kohut, it was a horror story.

Kohut is president of the Gallup poll, whose final New Hampshire survey was wrong by 17 points: it had put Dole ahead by 8; Bush won by 9. "I was dismayed," Kohut acknowledged yesterday.

This New Hampshire primary was perhaps the most-polled primary election in American history, and in the end, the Republican voters in the state confounded the predictions of nearly every published survey of voter opinion.

Many polls, including The Washington Post-ABC News Poll, caught the movement to Bush in the closing days, but only one -- the CBS-New York Times poll -- accurately foretold the final result.

Gallup's glaring error and the miscalls of other polling organizations once again raise questions about the accuracy of polls, their use by the media and the impact they have on voters' choices and the public perception of elections. In New Hampshire this year, news organizations' use of "tracking polls" to try to follow the movement of public opinion night after night came to dominate news accounts of the campaigning and the thinking of the campaigns themselves.

Tracking polls usually survey a relatively small number of voters every night: 150 to 400 in each party, in the case of The Post-ABC poll. The results are then averaged over several days. Pollsters believe this technique is helpful because it detects movement, but they also acknowledge its risks.

"If you're going to go into the tracking business," said pollster Peter Hart, "you better have a strong stomach and steady hands."

Richard Morin, polling director of The Washington Post, said tracking polls particularly can pose problems in a fast-changing race. He said they are "good at measuring movement" but not so reliable as predictors of the outcome.

In fact, it was tracking polls that showed the sudden collapse of Bush support in New Hampshire. The Post-ABC survey detected a falloff in the vice president's support during the two days before the Iowa precinct caucuses, a trend that continued after Dole defeated the vice president by 2 to 1 in Iowa.

Bush once enjoyed a New Hampshire lead of 20 points or more, according to many surveys of sentiment in the state. By late last week, according to the tracking polls, that lead was gone and the race was even; some surveys, including The Post's, found Dole ahead last Friday and Saturday. The Post reported three-day averages, and on that basis never had Dole ahead by more than 1 percentage point. By Sunday and Monday, The Post reported that Bush had regained momentum, though the last poll result reported in this newspaper showed the race dead even.

Because news accounts over the weekend were dominated by reports of Dole's success and apparent momentum in New Hampshire, the outcome made the Kansas senator look like a big loser in New Hampshire.

"Clearly, who wins in New Hampshire and what political consequences result therefrom are highly conditioned from expectations," said Dole pollster Richard Wirthlin, who also was caught by surprise. "Putting the race at 35 to 27 {Gallup's spread} had to create and reinforce the perception that we were far out in front." Wirthlin himself was telling reporters in New Hampshire on the eve of the voting that Dole would win the primary, though by then several media tracking polls showed the momentum had gone back to Bush.

Larry Bartels, a professor of political science at Stanford who studies the impact of voters' perceptions on elections, said it was "surprising that the perception -- the news -- was that Dole was catching up {when} the tracking was moving the other way."

Bartels said media accounts of how public opinion is moving in an election can affect the final choices of some voters -- the "bandwagon effect," he called it. Bush might have done better, Bartels said, if voters had better understood that the tide was running in Bush's favor in the closing hours of the campaign.

Voters who make up their minds at the last minute are the pollster's nemesis. Network exit polls of actual voters indicated that a significant number in New Hampshire did make very late choices.

The Gallup poll of 902 New Hampshire Republicans, with a 3 percent statistical margin of error, was based on telephone interviews conducted last Friday, Saturday and Sunday. Gallup's Kohut defended the accuracy of his poll as far as it went, but conceded that he made a miscalculation by cutting off voter interviews at 4 p.m. Sunday, before the League of Women Voters debate, with the available data indicating a volatile race.

The Boston Globe opted not to report on its own poll when the percentage of Republican undecideds jumped from 9 on Saturday to 19 on Sunday. "Our pollster was not confident," said Royal Ford, the Globe's national editor. "He never saw a burp that late before and he thought something was faulty."

Phil Balboni, news director of Boston station WCVB which broadcast the wrong Gallup numbers the night before the voting in New Hampshire, said it was a mistake to use those numbers more than 24 hours after the interviews were completed.

"We can't and will not do that again," Balboni declared. "I told Andy {Kohut}, 'For Super Tuesday, I want you to poll on Saturday and Sunday and Monday. We're not going to run polls on the 6 o'clock news based on Sunday."

In an apologetic memo to his clients distributed yesterday, Kohut wrote that "we did not put enough emphasis on {the softness of Dole's support} nor did we put enough emphasis on the fact that in primary elections voter sentiment can change literally overnight."

He added: "The lesson learned is that even if a front-runner's lead appears stable, it remains vulnerable to last-minute changes if support is soft . . . . As a consequence, we did not adequately address the possibility of a Bush victory!" He vowed to "do more analysis of the potential ways voter sentiment can change."Staff writer Maralee Schwartz and researcher Colette T. Rhoney contributed to this report.