The Washington Post reports that their new poll shows John McCain trailing Barack Obama and Hillary Clinton — a result that contradicts Rasmussen’s polling this week. Instead of showing a five-point lead for McCain, the WaPo survey has Obama up by 12. How can that be? It helps to have a huge oversample of Democrats and a lot of unregistered voters (via Jim Geraghty):

Sen. Barack Obama (Ill.) leads McCain, who captured the delegates needed to claim the Republican nomination Tuesday night, by 12 percentage points among all adults in the poll; Sen. Hillary Rodham Clinton (N.Y.) holds a six-point lead over the GOP nominee. Both Democrats are buoyed by moderates and independents when going head to head with McCain and benefit from sustained negative public assessments of President Bush and the war in Iraq.

Well, let’s take a look at that sample, shall we? It consists of 1126 adults, 15% of which are unregistered. That puts registered voters at 957. That’s enough to skew the sample, at the very least making its predictive value useless. Most reliable pollsters use registered voters for their entire sample, and the most predictive narrow that down to likely voters.

That’s not the only problem in this poll. Party identification breaks down 40/28/28 for Democrats, Republicans, and independents, respectively. Rasmussen’s party identification surveys show a much different breakdown for the population: 38.7/33.1/28.2, which halves the gap shown in WaPo’s survey. The survey then asks how the independents lean, and winds up with almost a 20-point gap between Democrats and Republicans — 55/36/7.

Didn’t anyone at the Post look at these numbers and wonder whether that reflected the nation as a whole, or perhaps a bias in sample collection?

The result is an unreliable poll, but one hell of a headline. McCain trails by twelve! And when McCain beats Obama or Clinton, we’ll hear once again that the election had to have been stolen — because all of the polls showed McCain behind.