By Howard Kurtz
Washington Post Staff Writer
Thursday, November 9, 2006; Page A39
The television networks were unusually cautious on election night, with several of them saying that their seemingly snakebit exit polls were close to useless.
When NBC projected at 10:54 that the Democrats would take control of the House, the call was based not on the network consortium of exit surveys, which its executives say tilted too Democratic, but on its own analysis of voting patterns in key districts. ABC called the House at 10:57, CBS at 11:01, CNN at 11:08 and Fox News at 11:17.
"We looked at the exit polls and noted what we thought was a pattern -- a 6 to 8 percent skew to Democratic candidates," Fox News Senior Vice President John Moody said yesterday. Once the surveys diverged from the raw vote totals trickling in, he said, "we decided not to use the exit polls in making projections. And because we want to be transparent with our viewers, we told them that."
Dan Merkle, director of ABC's decision desk, said Tuesday's polling caused "a definite concern. We once again saw the overstatement of Democrats that we saw in 2004." Merkle said he and his network disregarded exit surveys that were out of whack with their expectations based on preelection polling.
Others defended the effort. "We were aware of those weaknesses," said Kathy Frankovic, CBS's director of surveys. "We understood about potential problems in some places. We were comfortable using them the way we did."
Frankovic, who is also the spokeswoman for the National Election Pool -- the consortium of the networks and the Associated Press -- described the problems as typical of exit surveys, such as undersampling in bad-weather areas. She said she did not believe the samples were skewed toward the Democrats, but that further analysis would be required to provide a definitive answer. The exit poll was conducted for the National Election Pool by Edison/Mitofsky.
After Fox anchors criticized the polling on the air Tuesday night as being pro-Democratic, Joe Lenski, executive vice president of Edison/Mitofsky, said: "I know of no justification for Fox's statement."
The National Election Pool was reorganized and renamed after a virtual meltdown in 2002, when what was then called Voter News Service botched things so badly that the group pulled the plug and released no numbers. In 2004, what were widely acknowledged to be skewed polling samples had Democratic candidate John F. Kerry on his way to the Oval Office, which influenced some of the early on-air handicapping.
Even the most precise exit polls, of course, would not have enabled the networks (or the newspapers that also rely on the surveys) to call the Senate, because cliffhanger races in Virginia and Montana were still unresolved by morning. And there were no exit polls in the House races. So the networks were flying without their full instrument panel.
Despite the concerns about projections, all the networks used the polls to describe voters' attitudes.
Sam Feist, CNN's political director, said he had indications from his statisticians early on that the House was very likely to go Democratic, "but we wanted to be sure. We wanted to be cautious. There was some of the same Democratic oversampling we saw last time and we noted it. We told our staff, these don't mean anything. People forget, exit polls were never meant to call races."
Sheldon Gawiser, NBC's decision desk chief, said he found the polling "a little too Democratic," but not as far off as he initially thought before the magnitude of the Democratic wave became clear. "Anytime you have a set of poll results that aren't spot on, you're disappointed," he said.
Gawiser said he and other analysts have identified a fundamental problem with exit surveys from their own polls: "We certainly find that Democrats are more likely than Republicans to take them."
© Copyright 1996-2006 The Washington Post Company