In hindsight, they seem like the astrological predictions of supermarket tabloids:
* Democrat Harvey Gantt leads Republican Jesse Helms, incumbent U.S. senator from North Carolina, 47 percent to 41 percent in the last Charlotte Observer poll.
* Mr. Gantt leading 48 percent to 44 percent -- and picking up momentum -- in a recent Mason-Dixon poll.
* Mr. Gantt ahead 49 percent to 41 percent in an earlier Observer poll.
On Tuesday, Mr. Helms won a fourth term with 52.6 percent of the vote to Mr. Gantt's 47.4 percent.
So why were the polls so far off?
"I know what you're calling about," said Del Ali, vice president of Mason-Dixon Opinion Research of Columbia, Md. "Frankly . . . we had a very good day yesterday. We were right on the money everywhere except there."
Mason-Dixon polls correctly forecast many races, including tight gubernatorial contests in California, Texas and Illinois. But not the North Carolina Senate race.
The reason, according to Mr. Ali and other pollsters, lies with what's known as "the Wilder Factor" -- and what it means about the "undecided" vote.
The phenomenon takes its name from L. Douglas Wilder, Virginia's first black governor. Like other black candidates, including New York Mayor David N. Dinkins, Mr. Wilder squeaked out a narrow victory despite polls showing him with a strong lead.
Analysts say such discrepancies occur because some white voters don't tell pollsters the truth or simply declare themselves undecided. That can lead pollsters to underestimate support for a white candidate facing a black candidate such as Mr. Gantt.
"What's being termed the Wilder Factor . . . is becoming extremely frustrating in trying to analyze polls," said Joe Denneny, research manager for the Observer's polls. "It's almost like part of the electorate has a secret agenda."
Mr. Denneny said his polls were off by a similar margin only once before: in Charlotte's 1987 mayoral race between Mr. Gantt and Republican Sue Myrick, who is white. Mr. Denneny's polling organization also conducts The Sun Poll.
What appears to have happened with Mr. Gantt and other black candidates is that undecided voters -- most of whom are white -- all end up voting for the white candidate.
"Harvey, like other black candidates, did not receive any of the undecided vote in the polls," said Washington-based pollster Michael Donilon, who worked for Mr. Gantt and Mr. Wilder. "That was consistent with the Virginia election and the New York City election."
An analysis of Tuesday's returns suggests that that may indeed be what happened in North Carolina.
The final Observer poll, for example, showed Mr. Gantt getting 43.1 percent in the mountain region of North Carolina to Mr. Helms' 43.3 percent, with the rest undecided.
Mr. Gantt wound up with exactly 43.1 percent, while Mr. Helms captured 56.9 percent.
Along the coast, the poll gave Mr. Gantt 46.4 percent to Mr. Helms' 41.4 percent.
Mr. Gantt actually got 47.4 percent of
votes there to Mr. Helms' 52.6 percent.
In each case, the poll dramatically understated support for Mr. Helms, the white candidate.
"I think you just have to assume in elections involving white and black candidates that the undecideds break toward the white candidate," said Mr. Donilon.
Mason-Dixon's Mr. Ali said, "There are enough closet Helms supporters who were, number one, embarrassed to say they would vote for Jesse Helms, and number two, that they're not going to come out and tell you they're not going to vote for a black. That's what we found in the exit polling in Virginia."
Not that the so-called Wilder Factor was the only reason the poll results did not accurately project the outcome of the race.
Mr. Helms' final 10-day blitz on the campaign trail and over the airwaves eroded some of Mr. Gantt's crucial white support, Mr. Donilon said. Mr. Helms repeatedly criticized Mr. Gantt's character and past business dealings and accused him of favoring racial quotas that could cost white people their jobs.
Another factor in Mr. Gantt's defeat, Mr. Donilon added, was the unexpectedly high turnout.
Statewide turnout was 61 percent, compared to 50 percent in 1986, the last non-presidential election.
"One of the basic assumptions for a Gantt victory was that the vote would resemble a typical off-year election," said Mr. Donilon. "In fact the overall vote came closer to a presidential-year turnout, which meant that black voters, instead of contributing between 21 percent and 22 percent of the overall vote, probably contributed between 18 [percent] and 19 percent."
Jim Morrill covers politics for the Charlotte Observer, where another version of this article appeared.