Politico/GWU poll shows Romney, Obama in virtual tie

For the next few days, we’re likely to see polling on the presidential race that captures a moment caught in amber — before Mitt Romney picked Paul Ryan to be his running mate.  The first of these BR (Before Ryan) polls comes from Politico, with its Battleground poll conducted by George Washington University.  The poll shows almost no change since its last iteration in May, but it does have a curious outcome in one key demo:

The poll, conducted in the days leading up to the Ryan announcement on Saturday, finds that despite the unprecedented millions of dollars being poured into the contest and the non-stop attacks from each side, the top line numbers are essentially unchanged from a previous Battleground poll in early May.

Obama takes 48 percent of likely voters in the new poll, compared with 47 percent for Romney – a statistical tie and well within the margin of error. In May, the numbers were flipped: Romney was at 48 percent and Obama was at 47 percent. The poll found 5 percent of voters are undecided.

“The overall sort of broader scope of the ballot is that there’s been little to no movement,” said Republican pollster Ed Goeas of the Tarrance Group, who helped conduct the bipartisan poll.

The intro paragraph fails to note that it’s been mostly Obama who has poured money into contest.  Romney’s fundraising has been focused mainly on general-election spending, which left the summer months to Obama.  Despite a 3-1 spending edge, the race is pretty much at status quo ante.

Or perhaps worse:

Both candidates also maintained their previous advantages among specific demographic groups. Romney leads among independent voters by 10 points, 47 percent to 37 percent, the same margin he had in May. And Obama continues his advantage among female voters by 15 points – also similar to his May margin.

Er, how exactly can this race be tied when Romney leads indies by 10 points?  Obama won in 2008’s Democratic wave election (D+7) by seven points, and Obama won independents by eight points, 52-44, according to the 2008 exit poll.  Obama also won women by 13 points, 56-43.  How can he still within a virtual tie with Romney if he’s only gained two points among women but lost 18 points in the gap among independents?

The answer is exactly where you’d think it was — in the sample.  The Politico/GWU poll has a D/R/I of 36/29/34 for another D+7 election model.  If one thinks that the 2012 election will be similar to the 2008 model, that makes sense — but no one outside of DNC headquarters and MSNBC think that will be the case.  Even more odd, a follow-up question asks whether respondents vote strictly or mainly along party lines, or more evenly split it.  That presents a functional D/R/I of 39/36/20 [see update], which makes me wonder exactly where GWU surveyed for its poll.

If Romney goes into the election with a ten-point advantage among independents, this election won’t be close.

Update: Originally I had the functional split at 49/36/20, but I had it wrong; it should be 39/36/20, which defeats the point I was making.  That’s a reasonable split between Dems and Reps.