Poll Position

June polls of a horserace that ends in November aren't "reliably predictive," as the survey experts say. So why are Republicans so ballistic about new surveys, including Newsweek's, that show Sen. Barack Obama with a big lead?

The first and simplest answer is that Sen. John McCain needs to persuade—quickly—skeptical Republicans of all stripes that he has what it takes to beat a phenom with tons of cash. Even though it's four months to Election Day, McCain needs to generate some genuine confidence inside GOP ranks if he is to have a chance in November.

McCain has spent the last many days hammering Obama on energy policy, and his campaign is hoping to show some results from the effort, if for no other reason than to impress Republicans with money that the campaign knows what it is up to.

The conventional wisdom—based, in good measure, on the ongoing Gallup Tracking Poll—has been that the race is dead even and likely to remain that way for months, and that the early, long-distance exchanges between Obama and McCain have had little effect. That was my thinking, too, that the race would "break" late and take its true shape only after voters get to see Obama and McCain face off against each other in person.

Ironically, the Obama campaign is concerned about the polls from the other direction: They don't like the dead-even Gallup Tracking Poll. They have their own skeptical funders to impress, the ones who supported Hillary Clinton. So the Obama camp is focusing on specific numbers rather than the horserace per se. In a meeting with reporters today, David Plouffe, Obama's campaign manager, showed a PowerPoint centering on women voters, independents and red states that the campaign hopes to turn blue.

But two new polls—our latest NEWSWEEK poll and an even newer one from The Los Angeles Times and Bloomberg News—run counter to the conventional wisdom. They show Obama with wide leads, of 15 points and 12 points respectively.

The response to our poll was predictably nasty. Trent Lott, who left the Senate to become a Washington lobbyist, struck first. "I doubt the accuracy of that poll," he declared. "It is in Newsweek, after all." A day later, Karl Rove, George W. Bush's "architect" and now a Washington political consultant, chose one of the late George Carlin's seven dirty words to describe the quality of our poll in a private meeting with Republicans on the Hill (and he's an occasional contributor to Newsweek!).

Some of our competitors dismissed our survey, too, although they used a more polite, technical term: We were an "outlier." But then the LAT/Bloomberg numbers appeared yesterday, and the universe of "outliers" expanded. Maybe I missed it, but I haven't heard from Lott and Rove on this.

Instead, it was left to the McCain campaign to question the Times/Bloomberg poll in a study by its polling firm. Using dry, methodological language, the campaign made one central point, that the survey had had too many Democrats in it and too few Republicans: 39 percent Democrats and only 22 percent Republicans. No other recent poll had found that large a gap in a random survey of adults. (The Newsweek poll had a similarly large gap.)

But what if the two new polls were reflecting the reasonable possibility that, in recent days, both "strong" and "weak" Democrats had begun to move toward unity behind Obama? And what if, as economic fears deepen, gas and food prices rise and President Bush falls through the floor in terms of job approval, the GOP brand has become even more toxic than was thought possible?

There are some other, more technical, reasons why the Newsweek poll might differ from, say, the Gallup Tracking Poll. Every polltaker has his or her own theories of how best to do the job. Ours come from Larry Hugick, the chairman of Princeton Survey Research, who has been in the business since 1978, and polling for Newsweek since 1993. Unlike some firms, he explained to me, PSR always asks the "horserace" question first; his theory is that this ensures that answers aren't influenced by the wording of earlier questions.

PSR only calls landlines; Gallup calls cell phones, too. There is a big dispute among the experts as to whether this matters. In fact, according to Hugick, it may mean that Newsweek didn't sample enough voters under the age of 30, who've grown up with a cell phone in hand. If that's true, then Obama did as well as he did in our poll in spite of a cell phone handicap! We also poll in the evening; Gallup does so throughout the day. And our latest poll was taken over two days, June 18 and 19. Gallup's "tracking" poll is a "rolling average" over three days.

Even Gallup polls differ from client to client. Its own tracking poll has the race dead even; the one that Gallup does with USA Today has Obama ahead by six points.

All polls are snapshots of a moment in time, the cliché goes, and there is truth to it. My sense of our own Newsweek Poll is that it is very good at picking up—often ahead of the curve—early movements that other polls later validate. We are tuned to a high sensitivity, and catch quick, momentary reactions, some of which turn out to be prophetic, others of which turn out flat wrong.

A good example of the first kind was late last fall, when we were the first to show Obama ahead of Sen. Clinton among Democrats in Iowa. "We took a lot of heat for that one, too," Hugick recalled.

But nothing like in 1984. That was the most embarrassing example of the latter kind of survey. It predates PSR but remains seared into our institutional memory. The day the Democratic convention ended in San Francisco in 1984, the Newsweek poll showed Walter Mondale 18 points ahead of President Ronald Reagan. Mondale ended up getting clobbered, 49 states to one.

For readers and voters, says Hugick, the best way to judge the flow of polls is to look at them all. There are websites, such as RealClearPolitics.com, that constantly compile an average of all credible surveys. According to that average, Obama is ahead by 6.7 percentage points. Which means that, if we're still "outliers," so are the folks at Gallup.