Yesterday, I noted the topline results of the CBS/NYT poll, which showed Barack Obama and Mitt Romney in a dead heat — but which didn’t supply the internals of the poll.  Today, CBS follows up on their polling with an interesting and more nuanced look at the gender gap, in which Romney trails badly with single women — but not so much with married women.  In fact, those working and stay-at-home moms give Romney the edge:

One headline out of our poll is the shrinking gender gap. A month ago, President Obama had an 11-point lead over Mitt Romney among women voters. Today’s poll, taken after Hillary Rosen’s comments and the subsequent firestorm, puts the gap at six points.

But as everyone should know (and despite all we hear about the “female” vote), women aren’t some monolithic group. Our poll reveals sharp differences in opinion, for example, between married and single women.

Among MARRIED women, Romney leads Obama 49 to 42 percent. But among SINGLE women, Obama has a huge lead over Romney, 62 to 34 percent.

In another piece of bad news for Obama, the poll shows that Romney has more upside than Obama — which is exactly what you’d expect to find with an incumbent.  More than a third of respondents (37%) don’t know enough about Romney to form an opinion of him, while only 13% say that about Obama. (Really?  After more than three years as President?)  That is why having a dead heat at this stage of the race is such bad news for Team Obama, along with a re-elect number in the mid-40s.  Romney has much more room to improve his standing than Obama does, and that’s why Team Obama has been so desperate to attack Romney on a personal basis.

Finally, though, we did get a look at the sample and its partisan split.  What I can say without reservation is that it’s better than CNN’s triple-facepalm performance in turning a one-point edge among respondents into a weighted twelve-point Obama lead, but that’s damning with faint praise.  Once again, the weighting in the poll (specifically on registered voters) shows a curious sense of modeling which makes this much less predictable than their original respondent composition would have been.

Here is the original D/R/I of the poll, compared with the weighted D/R/I:

• General population: Original 33/29/38, Weighted 34/30/36
• Registered voters: Original 34/31/36, Weighted 34/26/40

This isn’t as dramatic as the CNN poll, but it’s still odd.  The weighting on the general-population responses seems unnecessary but harmless.  However, on the RV side, the weighting looks less harmless.  A D+3 sample at 34/31/36 would be a rational predictive model for the upcoming election, somewhere between the 39/32/29 of the 2008 election and the 35/35/30 turnout in the 2010 midterms.  However, the 34/26/40 puts Republicans at a significantly lower level (six points!) than even the nadir of 2008 while extending the Democrat advantage to higher than that election’s D+7, while overstating the impact of independents.  Why choose that model when the original survey’s composition would have sufficed?  Needless to say, that’s the question CBS and the New York Times probably won’t be answering.