New Hampshire: Graveyard of Pollsters

But one lingering question: Why did the final entrance poll estimates differ by so much from the final results. For example, the final estimate released from the National Election Pool put John Edwards support at 26 percent, and he ended up getting 32 percent. The John Kerry estimate was also noticeably different. NEP estimated 32 percent, Kerry got 36 percent.

Could it be that the very first survey by NEP was . . . wrong?

Right question, wrong answer, said Lenski, whose firm is conducting the NEP exit polls with Mitofsky International. Yes, the latest Democratic entrance poll had the biggest gap between entrance poll estimate and the final result ever since the networks started doing these surveys in 1984, he said. But it had nothing to do with the quality of the poll and everything to do with the quirky nature of the Iowa caucuses . . . and Rep. Dennis Kucinich (D-Ohio).

Here's what happened. Iowa Democrats caucused several times on Monday. The results of the first caucus, held shortly after 7 p.m. Central Standard Time, became the NEP final estimate; in this case, 26 percent for Edwards and 35 percent for Kerry.

But then Iowa Democrats began a furious half-hour of politicking and candidate-switching that ultimately ended in a final caucus tally that gave Kerry 38 percent and Edwards 32 percent support (technically expressed as a percentage of "delegate equivalents).

Lenski said his analysis of past entrance polls found that leading candidates "tend historically to gain two to four points in the delegate equivalents statewide," he said. Mondale in 1984, Gephardt and Simon in 1988 and Gore in 2000 gained 2 to 4 points in the delegate equivalents compared with the initial preference. That is also what Kerry did in 2004.

But what about that six-point surge for Edwards, a new record? He says that's the payoff Edwards got, thanks to a deal in which Kucinich pledged to shift he supporter to Edwards in precincts where he got less than 15 percent of the total support if Edwards would do the same for him in his low-performing precincts.

Needless to say, Kucinch, who got 1 percent support, had many more low performing precincts than Edwards.

"I think that is evidence that the deal with Kucinich helped to move a couple of points to Edwards in the delegate equivalents," Lenski said. "Also in our entrance poll Edwards was the top second choice of the relatively small number of Kucinich voters," more evidence of the K-Factor in Monday's caucuses.

Share your opinions with the Poll Watchers at polls@washpost.com