Poli-Sci Perspective is a weekly Wonkblog feature in which Georgetown University's Dan Hopkins and George Washington University's Danny Hayes and John Sides offer an empirical perspective on the issues dominating Washington. In this edition, Hopkins summarizes research showing that the auto bailout didn't have nearly the effect on the election that commentators thought.

In the wake of the first 2012 presidential debate, as former governor Mitt Romney gained ground in swing-state and especially national polls, a narrative took hold about the importance of the auto bailout. The Buckeye State was likely to decide the election, President Obama was over-performing in Ohio relative the country at large, and the Obama-backed bailout of Chrysler and General Motors was the reason why.

On Oct. 30, for instance, a Boston Globe article explained the logic: "as Obama campaigns for reelection, the controversial bailout has allowed him to secure Michigan and puts him within striking distance in Ohio, a critical battleground state where 1 in 8 jobs are auto-related." Like many campaign narratives, the narrative about the bailout's importance seems to have survived election night, with Politico running a story the day after the election entitled, "Auto bailout may have saved Obama."

Nate Cohn of The New Republic and Matt Yglesias of Slate have expressed skepticism, pointing out that Obama's margin in Ohio of just under 2 percentage points is consistent with its typical performance as leaning very slightly Republican.  Here, I want to add to that skepticism by looking at county-level vote shares alongside the location of Chrysler, GM, and Ford plants.  (Yes, I'm aware that Ford was not bailed out, but there's still an argument that the bailout helped keep Ford open as well, and Ford was certainly supportive of it.)

Let's start by considering the set of contiguous, largely midwestern states that have more than one Big Three auto plant.  On the left side, the map below shows those states, with the counties that have Big Three auto plants shaded black. The map on the right shows each counties' change in President Obama's share of the vote from 2008 relative to the nation overall. The redder the county, the sharper the swing away from Obama from 2008 to 2012.

At a quick glance, one might think there is a relationship. Counties with Chrysler, Ford or GM plants like Wayne County, Mich. (home to Detroit), Cook County, Ill. (home to Chicago), Hamilton County, Ohio (home to Cincinnati) and Cuyahoga County, Ohio (home to Cleveland) all saw relatively small movements of less than a few percentage points toward the GOP. But those counties are not just home to auto manufacturing--they are also large urban centers with many black voters.

In auto-making counties that don't meet that description, the story was different.  In Howard County, Ind., for instance, Obama's vote share declined from 46.0 percent to 41.7 percent in spite of the county's two Chrysler plants. In St. Charles County, Mo.—the site of GM's Wentzville Assembly plant—Obama's vote share declined sharply from 44.6 percent to 38.7 percent. So it seems likely that something other than the auto plants themselves might be at work.

The figure below bolsters that case by plotting all county-level vote shares in 2008 and 2012, with counties that have auto plants listed in red using the letters of their state. Counties above the 45 degree line become more pro-Obama from 2008 to 2012, while the majority of counties below it became more opposed.  There isn't much evidence that the counties with auto plants were very different from all U.S. counties--those that were strongly supportive of Obama remained so, and those that were not did not swing into his camp.

To separate out other key factors, I then turned to regression, a statistical tool.  If we account only for the state in which each county is located, we would estimate that local auto plants are associated with a 1.7 percentage point increase in Obama's vote share in the set of midwestern states with more than one plant from an American automaker, or a 1.5 percentage point increase in all continental U.S. states outside of New England.  

Once we account for the county's population and its basic ethnic/racial demographics, however, these estimates decline to between 0.3 percentage points and 0.5 percentage points, and the associated uncertainty makes them indistinguishable from zero.  The results are quite stable when we introduce additional controls for demographics, local industry and the volume of political advertising. Counties that are home to Big Three plants just don't behave very differently from similar counties that are not.

Elected officials, candidates and commentators alike are quick to emphasize the alleged role of self-interest in voter decision-making.  When Romney said in the first debate that “if you are 60, or around 60 or older, you don't need to listen any further” to his Medicare proposal, he was appealing to current seniors' self-interest.  So, too, was Obama, when we responded that “if you are 54 or 55, you might want to listen.”

But decades-old literature in political science suggests otherwise, demonstrating that voters are only rarely moved by their self-interest, at least narrowly understood.  Now, that's not to say that self-interest has no role in politics—the behavior of organized interest groups strongly suggests otherwise.  But the evidence from the auto bailout reinforces what we have seen in health care and in many other cases: the direct influence of self-interest in voter decision-making in any one election is surprisingly small.