One thing to keep in mind as you peruse those latest Wisconsin and (especially) Ohio and Texas polls is that African American support for Obama continues to be implausibly low. We've seen similar numbers leading up to primaries in South Carolina, Georgia, Maryland, and Missouri--all places where Obama attracted in the neighborhood of 80 percent of the black vote (and often closer to 90) on Election Day.
This is a big deal, since Ohio and Texas have significant African American populations--14 and 21 percent of the 2004 Democratic primary electorate, respectively. The difference between a merely strong showing among African Americans and an overwhelming one can be several points overall. For example, consider this recent Quinnipiac poll in Ohio showing Obama up 64-17 among African Americans. If, as seems plausible, the final number ends up closer to 85-15, where it was in nearby Missouri, and the black share of the primary electorate ticks up a few points relative to 2004 (as it consistently has), we could be talking about an overall difference of three or four points. The effect would be closer to four-to-five points in Texas, where this recent poll has Obama's lead among African Americans at 65-16. Those are pretty significant differences.
Why do polls seem to consistently understate Obama's African American support? I can think of a number of plausible explanations--for example, maybe certain black voters are undecided until the sheer energy of Election Day (or the actual prospect of voting for a viable black candidate) hits them in the face. But there's also compelling evidence for a reverse Bradley effect--i.e., that a significant number of black voters are, for whatever reason, just misleading pollsters.
--Noam Scheiber