Do ‘undecideds’ break white?

That’s the argument of 1996 Republican National Convention manager Bill Greiner. He surveys several races of the recent past with African-American candidates, both Republican and Democratic, and including winners and losers.

He says the anecdotal evidence is that undecideds break sharply white at the end, and so, Obama needs to break 50 percent.

His case seems compelling enough for further discussion. Or, more examination of its flaws.

First, I was suspicious with the claim that Obama "needs to" do anything being made by a Republican. This seems like a "raise the bar" claim being made by a surrogate for McCain.

And, as for discussion, over at FiveThirtyEight, Nate gives him real analysis beyond just discussion.

Cherry-picking individual polls instead of poll averages. Cherry-picking contests, including excluding the Ed Rendell-Lynn Swann Pennsylvania gubernatorial tilt.

Or, excluding the Obama-Keyes race in 2004 (one that Nate doesn't mention, either); by Greiner's logic, undecideds should have stayed at home. He says nothing about voter turnout there.

Shorter Nate? Greiner did some big-time