). "Some are good, and some are junk."
So how do you sift through this numerical jumble? Here are a few hints:LOOK FOR TRENDS. Individual polls are often unreliable because of their margin of error, usually plus or minus three percentage points. That means two polls finding candidates in a dead heat at 45%-45% or at 52%-42%, as happened in late September, could be telling the same story. But if polls show just one candidate gaining, opinion is almost certainly moving that way. For example, George W. Bush trailed John Kerry in 14 of the 24 national polls in August, with the Prez ahead in 6 and tied in 4. But Bush had the upper hand in 28 of the first 33 September surveys, vs. just 1 for Kerry. The rest were tied. Undeniably, the momentum was with Bush.BEWARE THE OUTLIERS. One rule of media coverage: The bigger the shift in a single poll, the louder the buzz. So a Sept. 13-15 Gallup survey giving Bush a 14-point lead among likely voters got a huge amount of publicity, even though 7 of the 8 other polls conducted that week showed a margin less than half the size. One way to figure out the real state of play is to scan the most recent polls. (A good Web site is realclearpolitics.com.) If you take the 10 most recent surveys, eliminate the extremes, and average the rest, Bush led Kerry by 6.3 points on the eve of the debates. This "poll of polls" isn't scientifically valid, but it is a good unofficial cheat sheet.CHOOSE YOUR POISON. If you think most '04 polls are stacked against Dems, Zogby is the poll for you because it adjusts its numbers to account for party affiliation. GOP partisans will like the Fox News/Opinion Dynamics Poll, whose results have generally been more favorable to Bush. Closest to the statistical center is the Associated Press/Ipsos Poll. (Its survey released on Sept. 26 gives Bush a seven-point lead.) Once you pick the one you like best, stick with it: Trends are more important than the snapshot of a single poll.SILENCE THE SPINNERS. It's a good idea to ignore partisan talking heads. For example, Kerry campaign manager Mary Beth Cahill sent an e-mail on Sept. 24 saying that "five new national polls this week show the race tied." Actually, 10 new polls showed Bush ahead, but 5 were within the margin of error.FORGET POLLS. Besides data overload, there are two new reasons to ignore surveys entirely: They could be missing the real picture because most measure "likely" voters and might not reflect a recent surge in registrations. And they could be flawed by failing to take into account people who either routinely screen their calls or have abandoned land-line phones. Republicans were none too happy when Hollywood tapped a prominent Democrat, former Kansas congressman and Clinton Cabinet member Dan Glickman, to take the helm of the industry's high-profile lobby shop. Now revenge could be at hand. Republicans are talking of torpedoing legislation coveted by the Motion Picture Association of America. The bill, which would make camcording films a felony punishable by up to six years in prison, passed the House on Sept. 28 and should be on its way to final negotiations with the Senate. But GOPers are talking about deep-sixing that provision. It's the only legislation Republicans can kill to hurt the MPAA without causing too much collateral damage to the more obedient Recording Industry Association of America, which last year hired a well-connected Hill Republican -- Mitch Bainwol -- to fill its top job. In a nation fundamentally divided along cultural lines, the factor most closely tied to Presidential preferences may well be whether a voter is a born-again Christian. President George W. Bush leads John Kerry in all 16 states where 30% or more of the population is fundamentalist Protestant, according to a BusinessWeek analysis of data released on Sept. 27 by the National Annenberg Election Survey. Of eight states where evangelicals account for less than 20% of residents, the Democrat leads in six. (Heavily Mormon Utah and strongly Catholic Wisconsin are the exceptions.) Fourteen small-population states were not included in the results.