Yesterday, the Washington Post published a chart showing a “compelling” relationship between higher turnout and President Obama’s performance last November. The implication was that Obama’s turnout operation was pretty central to his reelection campaign. I happened to be at the Atlanta airport when I first saw this, so I wasn’t positioned to write a quick response (although that didn’t prevent some serious eye rolling). Fortunately, Alan Abramowitz, a political scientist at Emory University, was thinking the same thing about 20 miles to my north, and he convinced the folks at the Post that turnout wasn’t as significant as the chart suggested, since change in turnout only had a very negligible effect on the change in margin. That’s consistent with more rigorous research from other political scientists, like Seth Masket, John Sides and Lynn Vavreck (whose forthcoming book you’re going to want to read), and others that I’m probably forgetting.
But Abramowitz decided to go an extra step further, saying:
“Turnout declined almost everywhere [in 2012 as compared to 2008] but the smaller the decline, the greater the drop in Obama’s margin [...] But it’s a very weak relationship."
Well, “very weak” is an understatement.
The entire relationship results from two outlying data-points: Alaska and Utah, the two lonely dots in the upper left and lower right quadrants.
And those data-points aren’t just outlying, they’re unrepresentative. Alaska was Sarah Palin’s home-state; Utah may as well be Mitt Romney’s. So those two states aren’t really telling us much about the relationship between turnout and Obama’s performance, they’re probably just reflecting something else entirely: That Obama did better in Alaska once Palin was off the ballot; Obama crashed in Utah once a Mormon was on the ballot.
What happens when you take those states out? The relationship between higher turnout and Romney’s strength evaporates. In fact, it flips: Now Obama does best in the states with higher turnout.
You get a similar result if you exclude Wisconsin and Arizona, as well.
Of course, this relationship is so, so weak that you should *not* cite it as evidence that Obama's turnout operation mattered. There's real research that addresses that question. The point is that if there is a relationship between the decline in turnout and Obama's margin, it's probably the opposite of what Abramowitz suggested.