On Oct. 25, less than two weeks before last Tuesday's election, The N&O published a poll suggesting that the proposed Wake County school bond referendum would fail. The poll found that 54 percent of voters surveyed said they would vote no, compared to 35 percent voting yes and 11 percent undecided.

On Election Day, the referendum passed, 53 percent to 47 percent.

Wha' happened?

That's not yet clear, but there are red faces in The News & Observer newsroom and at the offices of its Maryland-based pollster, Research 2000.

"It's like getting a story wrong," said Van Denton, The N&O's metro editor. "It's embarrassing and it hurts the credibility of the paper." Denton said he hoped readers who used the poll also informed themselves, before voting, by reading the extensive news coverage that The N&O provided.

The poll also was sponsored by WRAL-TV, which mentioned it regularly in coverage of the bond referendum.

Del Ali, president of Research 2000, blamed the bad numbers on the timing of the survey. The poll was conducted Oct. 19 and 20, nearly three weeks before the election. In politics, he said, "that's an eternity; it really is." He said he's confident that the numbers were right in that third week of October, that the bonds would have lost if the election were held then and that voter sentiment moved that much before the election.

Denton said The N&O chose to conduct its poll and publish the results early so as not to influence the outcome of the election. But several interested parties whom I talked to think the N&O poll did just that.

Francis X. De Luca, a leader of the anti-campaign, said he thought the results lulled his supporters into complacency and energized the bond proponents to work harder in the last two weeks. De Luca agreed with Ali that the findings were probably accurate at the time the poll was taken, but that the proponents' sprint in the final stretch, combined with low turnout, changed the outcome.

Justin Guillory, with the Raleigh firm Public Policy Polling, said he also thought the N&O poll influenced the outcome. "People trust the N&O," he said. "When they saw it was losing, a lot of people really believed that."

Guillory's firm, which is independent but works only for Democratic candidates, conducted three polls on the bond referendum, and all three showed the bonds winning. He expressed chagrin that The N&O didn't give much attention to those findings.

Other polls conducted by bond supporters -- Friends of Wake County and the Wake County Homebuilders Association -- also showed the bonds winning. The N&O noted that in its poll story.

Research 2000 and The N&O also whiffed on an Orange County referendum over whether to add two members to the Board of Commissioners. The poll showed the initiative losing, 37 percent yes to 56 percent no, with 7 percent undecided. But the referendum passed overwhelmingly, 69 percent to 31 percent. Ali, of Research 2000, said it's more difficult to get voters to respond knowledgeably to questions about a referendum, which may have complex nuances, than to a choice between two candidates. The referendum question in Orange County was about 500 words long.

The poll did get it right on the Durham district attorney race. The poll showed Mike Nifong winning with 46 percent of the vote. He won with 49 percent.

Polling is as much an art as a science, polling professionals will tell you, and the results can be affected by factors such as size of the polling sample, the margin for error and how questions were worded. The N&O actually had a larger sample -- 600 likely voters -- and a smaller margin for error -- 4 percent -- than the bond proponents' polls. Denton said Research 2000's results in the past were reliable.

But there was some Wednesday-morning quarterbacking at the paper about other factors, specifically about how the likely voters were selected for the sample and about the wording of the bond question (it didn't tell respondents much about the pros and cons).

All that will be fodder for discussion for future poll stories, but it illustrates that there is plenty of opportunity to get polls wrong. Given that polls likely can influence outcomes, the question is whether newspapers should even be involved in such an unscientific science. It does put our reputation at stake.

Denton, the metro editor, maintains that polling is a proper role for newspapers. "During a debate over important issues or hot political races, we turn to a poll as a scientific way of answering the questions that are under debate in the community," he said. Newspaper polls can provide an independent check on the polling results put out by the partisan sides in an election to bolster their candidates.

I agree that polling is a valuable tool to give you important information about issues in the community. In that respect, it's not that different from investigative reporting or any other enterprise journalism. As long as the polling is done carefully and responsibly.

In that Oct. 25 story about the Wake Bonds, Ali, the N&O's pollster, was quoted as saying he would be "shocked" if the bond request were approved. Yes, he told me Friday, he indeed was shocked, even though he was confident his results were accurate.

But he added, "that's one I'd like to have back."