Obama by 3? Or by 12?

It is impossible to reconcile the new LATimes/Bloomberg poll, showing Barack Obama leading by 12 points -- 49% to 37% for John McCain -- with the Gallup daily tracking poll, which has show Obama's lead in the 2%-3% range each of the past five days.

Disregarding Gallup and other recent polls (Rasmussen, Fox, Zogby, ABC) showing the margin around 5%, David Weigel of Reason magazine exults over the anomalous LATimes result:

The country desperately wants a Democratic president at the moment.

Even if this was what the poll showed (49% is not "the country") voters will not have the choice of voting for a generic Democratic president -- or against a generic Republican. In 1996, I was among those who believed that any Republican could beat Bill Clinton. Unfortunately "any Republican" wasn't on the ballot; Bob Dole was. He got 41%.

One reason that Weigel believes that "the country" is so wild for Obama is because Weigel is college educated, under 30, single, living in Washington, D.C. -- thus, a member of several key Democratic constituencies. Weigel inhabits a millieu of Obamaphiles. I'm reminded of Pauline Kael, who expressed amazement in 1972 that Nixon had won, since no one she knew had voted for him.

I'll keep watching polls with a jaundiced eye. Meanwhile, guess who said this:

A lot of times, media polling, if it shows that a candidate is not doing that well, it might, hurt them getting money for their campaign, but that's the name of the game.

That's Susan Pinkus, director of the LATimes poll, in a 2003 interview with PBS.

UPDATE: Allahpundit looks at the internals:

[W]e’ve got the same lopsided sample here that we had in that 15-point Newsweek spread last week — 39/22/27 among Democrats, Republicans, and independents, respectively. Even so, the LAT’s sample was random, and even Gallup concedes that Dems now lead the GOP 50/36 in party identification. We keep dismissing these polls where the leftist contingent vastly outnumbers the right, but, er, should we?

It's a good question, and I'm sure that Gallup and other pollsters are going to take a close look at their methodology in the wake of these two recent off-the-chart results. If Gallup and others have been "weighting" their samples to approximate a past pattern of party parity that no longer exists, then perhaps the LAT/Newsweek double-digit lead is real.

But I still doubt it. If you look at the actual primary results, Obama faded after March. He lost Pennsylvania, Indiana, West Virginia and Kentucky because blue-collar voters preferred Hillary. Therefore, I am suspicious . . . .

WHOA! Holy freaking cow! I just checked the crosstabs of the LATimes poll:

18-29.......20%
30-44......32%
45-64......36%
65+.........12%

Right there's your problem, Mister! Retirees only 12 percent of the sample? Compare this to the 2004 exit polls:

18-29......17%
30-44......29%
45-59......38%
65+..........16%

LATimes has oversampled younger voters and undersampled the crucial geezer constituency. They don't break out voter preferences by age group, but you can be sure that McCain does much better with voters 65-plus than he does with under-30s.