About that 47-47 Poll

Republicans are touting the National Journal poll showing the race tied at 47. They're especially pippy about it because it's tied even though the sample is +7 Democratic. But there's something else in the sample that I wonder about.

The poll's respondents were 74 percent white, 11 percent black, and 8 percent Latino. Note that the 2008 electorate, according to exit polls, was 74, 13, and 9. 

Ruy Teixeira, the leading electoral demographer whom I've cited many times, thinks it's more likely that this year's white vote will be 72 percent. Maybe 73. Says he'd be surprised to see 74.

Will black turnout really drop 2 percent? I have a hard time seeing that. Maybe. Some smart people think it might, that 08 was historic and all that. But I think African American voters are pretty revved up, myself.

Polling demographics, unlike party ID, are a legitimate basis on which to critique polls. As I explained last week, pollsters can't stack or dictate party ID. All they can do is ask people. But they can, and do--it's their job--dictate their sample's demographic makeup. I find it interesting that the Journal people--and again, maybe they're right--think black turnout will drop by 2 percent. That wouldn't be my hunch.

Anyway, Nate Silver has Obama up 3.9, which smells about right to me. 

UPDATE: A smart tweeter noted that a drop of 2 points in this context means a drop of 15 points in real terms. It's like this. Around 130 million people voted last time. If black turnout was 13 percent of that, that's 16.9 million. Assume 130 million again, just to make the math easier. Eleven percent of 130 million is 14.3. The difference of 2.6 million represents a decrease of 15 percent in black turnout. Really??