Oversampling in the Post/ABC Poll?

Update:: Now I see there's little question about oversampling, see *note below. But according to a text box, final results were adjusted to reflect national racial percentages.

Via a Newsbusters post about surely benign headline juxtapositions, Tim Graham links to an Ankle Biting Pundits item on a Post/ABC poll:

In this story (posted Monday, Sept. 12th at 5:31 p.m.), here's what we're told about the poll respondents:

Quote:

A total of 1,200 randomly selected adults, including an oversample of 200 African-Americans, were interviewed Sept. 8-11 for this survey. Margin of sampling error for results based on the full sample is plus or minus 3 percentage points.

Yet, in this write-up of the same story (posted September 13) the oversampling of blacks in not mentioned.

Quote:

The poll was conducted with 1,201 randomly selected adults, in an interviews between Sept. 8-11. The margin of sampling error is plus or minus three percentage points.

True. The second story, by Richard Morin and Michael Fletecher,  is on page A8 today; the first, as often is the game plan for Post/ABC polls, is the initial online teaser, by Morin.

I'm sending an email to Morin. One good thing washingtonpost.com does is link to some of the polling demographics. There's just one racial breakdown at the link, shown at left. I think it's combining total Hispanic self-identification with everyone else's ID, so that we end up with the breakdown for all racial categories in the poll. But we end up with 11% black Americans, which would be a slight undersample of the 12.3% measured in this 2000 U.S. Census Bureau brief.  Yet the first story, by Morin, mentions oversampling of blacks. If I read Morin correctly, they interviewed 200 African-Americans, about 16.7% of the total being polled.

So basically I have no idea.

The oversample is important because of typical party affiliations among black Americans and because of the following, linked here from the first story by Morin but mentioned in both:

Attitudes toward Bush and the government's overall response to Hurricane Katrina fracture along the twin fault lines of race and partisanship. Nearly three in four whites doubted the federal government would have responded more quickly to those trapped in New Orleans if they had been wealthier and white rather than poorer and black. But an equal share of blacks disagreed, saying help would have come sooner if the victims had been more affluent whites.

*Aha! The mention of an oversample is in a footnote on graphics printed over at the jump on A9. But Ankle Biting Pundits couldn't have known that, since it's not included online (though all the graphics printed in the paper are). My email to Morin will now include my question about demographic data online and why they chose to oversample black Americans. The footnote, which is not being called up with the story as of 1:15 pm, includes the following comment:

Final results were adjusted to reflect the country's actual racial distribution.

I've emailed Morin about what problem such over-sampling is intended to address, among other questions.