atthew Robinson is writing a book about the dangers of polling. He is the 1999 Phillips Foundation Journalism Fellow.

Going into Labor Day Weekend pundits and reporters covering the presidential election were shocked, appalled and thoroughly baffled. For weeks the importance of a Labor Day lead in the polls had been hyped by nearly every media outlet. But three polls (Research 2000, ICR and Gallup) put the race into a statistical dead heat.

Enter Newsweek, coming to the rescue of the working press. In a forth poll, Newsweek discovered a tremendous 10-point lead for Gore. And while this poll gave the apoplectic talking heads a well-needed breather, it also severely deceived the public. A new Reuters/Zogby poll gives Gore a more modest six-point edge.

Newsweek's polls are deceptive because employs sloppy methods. But it is more misleading because of its immediate results: Giving journalists exactly what they want. The following articles focused on the nearly unbeaten track record of candidates who were ahead Labor Day weekend.

Journalists love such trivia, which is why polls are so deadly in their hands. They follow the polls with all the vigilance they once reserved for policing government and exposing errant politicians, even changing their coverage as a candidate rises or falls in the all-powerful measures of public opinion (see sidebar).

Every single one of the Sunday morning talk shows cited the Newsweek poll as the new numbers became the subject for debate. Anxious pundits used the numbers to offer their explanations for why Gore is up and Bush is down. The explanations ranged from the success of Gore's convention speech, to his choice of Lieberman, to missteps by Bush and Cheney.

But is the Newsweek poll accurate?

Simply put, if there was ranking system for polls out there, Newsweek would rank at the very bottom. In the world of cars, Newsweek's polls are a 1975 Gremlin: clunky, technologically primitive, and a vehicle for liberals.

Newsweek's polls have consistently given Gore the highest support among the dozens of polls over the last few months. They're numbers shouldn't just be suspect. Because of Newsweek's consistent use of registered voters and low samples (less than 600 people in one case), the magazine stupidly or cunningly sets itself up as propaganda mill for the Gore campaign, sacrificing accuracy and objectivity, to join the mad dash for polls.

On August 18, Newsweek had Gore leading Bush by six points (48 to 42 percent). A voter.com poll put Bush up by five points just two days earlier. And on August 19, a USA Today/CNN/Gallup poll had Gore leading by only one point (47 to 46 percent).

In early August a Reuters/Zogby Poll had Bush up by 17 points (August 6). Newsweek: Bush up by just 11 (August 4).

Newsweek glides through using methods that are misleading and should be considered utterly useless by responsible journalists and vigilant citizens.

Newsweek continues to use registered voters. Since 1996, journalists should have learned the perils of using such weak samples. Registered voters are very different from likely voters. They are less informed, less likely to go to the polls, and also more Democratic.

The result is a more liberal and often very misleading result come Election Day.

Unreflective journalists, however, don't care about such methodological tripwires. They blunder along searching for any and all poll numbers so they can write about the crowd's reaction to the Game of politics. They offer up their explanations and spin their stories using the patina of public opinion with nary a thought for a poll's accuracy.

Newsweek isn't the only one to use botched methods.

Readers of USA Today got a quick and decisive lesson in just how ridiculous the media drive for poll numbers can be. On Monday August 7, 2000 the paper blared in block letters: "Bush Lead on Gore Grows to 17 points." The USA Today/CNN/Gallup Poll followed on the heels of the Republican's very successful convention — so it was easy to understand why Bush might be doing well. But the very next day another front-page story reported that Bush's lead had shrunk to just two points after Democratic presidential candidate Al Gore chose Sen. Joseph Lieberman of Connecticut as his running mate.

Why the discrepancies in the numbers? Was Lieberman that popular? Was Lieberman's condemnation of Bill Clinton decisive as some pundits reported? For anyone familiar with the level of ignorance of the average voter, there is little chance voters had even heard of him. In fact, more than half (53 percent) of all registered voters said they did not know enough about Lieberman to rate him, including 24 percent who admitted that they had never heard of him at all. (These numbers didn't make it into the story.)

The reason for the huge drop in Bush's lead was methodological. In fact, it was the same wrong-way Corrigan polling used by Newsweek. In the survey that showed a 17-point lead, the polling firm Gallup interviewed likely voters. In the second poll, Gallup interviewed registered voters who tend to be a far less accurate measure.

In its press release Gallup noted: "Typically, although not always, likely voters tend to be more Republican than registered voters are, giving Bush about a 4- to 5-point larger lead."

USA Today neglected to report this part of Gallup's work. Newsweek also keeps this information under wraps. When the USA Today/CNN/Gallup poll returned to likely voters later in the week, Bush was back up 16 points.

The difference between Newsweek and USA Today, however, is that Newsweek insists on this methodology known to bias toward liberals like Gore. Is this incompetence? Or just laziness?

In the end it doesn't matter.

Newsweek polls should be ignored as reliable measures of a candidate's electoral strength. Period. Journalists may not be able to resist the temptations of bad polling. But citizens should start insisting on the use of likely voters. It may cost media pollsters more, but in the end it means a more accurate basis for public debate than what is offered today.