Lots of poll data to peruse today, but let’s start with the Washington Post/ABC tracking poll.  For the third straight day, it reports that Mitt Romney has a one-point lead over Barack Obama, 49/48 — but its internals tell a different story.  And one reason for that is its shifting sample, which has gotten significantly more Democratic.

The Post reports the topline numbers:

For the third consecutive day of the Washington Post-ABC News national tracking poll, a single — statistically insignificant — percentage point separates the two presidential contenders: 49 percent of likely voters back Republican Mitt Romney, and 48 percent support President Obama.

The parity in the contest shows up elsewhere as well: the two candidates are just two points apart when it comes to dealing with taxes, and they are three points apart on health care policy. The poll’s margin of sampling error is plus or minus 3.5 percentage points for the sample of 1,278 likely voters.

Just outside the error margin — albeit barely — is Romney’s five-point edge on issue No. 1, solving the nation’s economic troubles. Romney’s advantage here peaked at nine points late last week (52 to 43 percent); now, the split is 50 for the former Massachusetts governor to 45 percent for the president.

Also at a slender five points, Obama ekes out an advantage on understanding people’s economic problems. This had been a point of regular advantage before Romney appeared to neutralize it earlier in the tracking poll. In the new Washington Post Virginia poll, Obama holds a 13-point lead in economic empathy, consistent with a mid-September Post poll in that key battleground state.

What’s not statistically insignificant is Mitt Romney’s lead among independents.  He wins those 55/40, an overwhelming advantage in a group that Obama won by eight in 2008 on his way to a seven-point win overall in a big Democratic turnout.  That’s a 23-point shift in the gap among roughly 30% of the electorate.

One more takeaway from this poll is the gender gap.  In 2008, Obama won the gender gap by 14 points, with a +13 among women and +1 among men.  In this tracking poll, Romney has a 2-point advantage in the gender gap, with a +10 among men and a -8 among women.

If Romney wins independents by 15, and the gender gap by 2, how can he only be leading by one overall? Simple — the Post/ABC sample has a declining number of Republicans in its sample.  Today’s rolling three-day average has a D/R/I of 35/28/34, which would put Republicans at seven points below their 2010 turnout and five points below their 2008 turnout.  Two days ago, the sample was D+4 at 34/30, but now it’s D+7.

The toplines are simply not credible in relation to the internals.  If Romney is winning independents outside the margin of error and has neutralized the gender gap, Obama can’t possibly be within one point overall.  Either the toplines or the internals are way off.  Even the Post’s Chris Cillizza wondered at the disconnect:

President Obama has a problem with independents. And it’s not a small problem.

In the last three releases of the tracking poll conducted by The Washington Post and ABC News, Obama has trailed former Massachusetts governor Mitt Romney among independent voters by between 16 and 20 percentage points. …

So, what gives? Why is Obama — at least according to the Post-ABC data — having so much trouble with independents?

The answer lies in the fact that most independents are not, well, independent. Of all the likely voters who called themselves independents in nine days of the Post-ABC tracking poll, fully three-quarters (75 percent) — said they tend to lean toward one party or the other. (The remainder are known as “pure” independents.)

And it’s among those shadow partisans that Obama is struggling. Ninety-two percent of Republican-leaning independents said they plan to support Romney, while 84 percent of Democratic-leaning independents are backing Obama.

Except, of course, that this dynamic has always applied to self-described independents, in every election.  The comparison cycle on cycle remains entirely valid, which is more than what one can say for the Post’s shifting sample.