Which voters’ Presidential preferences changed from 2008 to 2012? That straightforward question is central to making sense of yesterday’s election outcome–and yet it is a deceptively hard question. For example, consider the results from this year’s exit polls. They indicate that Hispanic voters backed President Obama by a margin of 71 to 27, up from his 67 to 31 victory in 2008. But how much of that shift is due to the changing composition of the Latino electorate, and how much stems from preference changes among Latinos who voted in both elections? That question has important political implications–and yet, standard exit polls cannot answer it.
Fortunately, a team of researchers at the University of Pennsylvania and Georgetown University is conducting an ongoing panel study of American adults that can help answer such questions. Last month, this Institute for the Study of Citizens and Politics survey used GfK’s probability-based online panel to recontact just over 2,600 Americans who had registered their vote preference during the 2008 general election campaign. Here, I use the panel to look at the shifting nature of the two parties’ coalitions.
Of the 1,092 respondents in the sample who backed President Obama during the 2008 general election campaign, 925 continued to support him in October 2012, meaning that he retained 85% of his supporters. Another 6% backed Governor Romney, while 8% weren’t sure. At first glance, that would seem to be a sizable defection rate. But now consider what became of Senator McCain’s 964 supporters in the panel. 790 backed Romney four years later, for a retention rate of 82%. Nine percent were now with President Obama, and 6% were undecided. In other words, McCain supporters and Obama supporters in 2008 switched their candidates in 2012 at roughly similar rates. A few caveats are in order. First, this is a sample of adults (rather than registered or likely voters), and is subject to panel attrition and other biases. Also, since Obama originally had a larger number of supporters, similar retention rates between Romney and Obama are compatible with a tightening race in 2012. Still, those figures make it clear that not all of the movement over the first term of the Obama administration was in the GOP’s direction, a point that is easily missed in much of the election coverage.
These panel data allow us to examine the characteristics that make prospective voters more or less likely to defect from their 2008 candidates–and since we are re-interviewing the same people four years later, we can be sure that changes in the composition of the electorate aren’t what’s really driving our findings. In the figure below, I plot the estimated effect of various demographic indicator variables on the probability that a 2008 McCain support backs Romney in 2012. I use the base of the arrow to indicate the probability for the reference group, and the head of the arrow to indicate the probability for the group in question. A leftward arrow indicates a factor that makes people more likely to defect from the GOP, while a rightward arrow indicators a factor that makes them more likely to remain in the GOP camp.
Strikingly, Hispanic adults who supported McCain stick with Romney only 65% of the time in this model, while non-Hispanic McCain backers are with him 84% of the time. The relatively low level of Hispanic support for Governor Romney did not simply come from the changing composition of the electorate, but from changes in preferences among Latino adults as well. Beyond that, it is noteworthy that older voters were more likely remain with the GOP candidate across the two elections. Romney’s retention rate for voters over 64 was 89%, even given all the campaign discussions of the Ryan Plan and of transforming Medicare. Age is emerging as an increasing demographic cleavage between the two parties, a point which deserves its own blog post (and more). Those making over $50K annually are also more likely to stay in the GOP camp (at 86% versus 79%), suggesting a growing cleavage based on income as well. Notice, too, that residence in one of the nine contested swing states leads to a decline in Governor Romney’s retention rate, from 84% to 77%.
We now consider the same figure for President Obama, shown just above. African American Obama supporters in 2008 are more likely to be with the President again in 2012, with his retention rate among Blacks at a hefty 93%. Those who report incomes over $20K are markedly less likely to stay with the President (82% versus 95%), again reinforcing that income played a role in the shifting coalitions of GOP and Democratic support. As was the case for Governor Romney, registered voters are less likely to defect from their 2008 choice. But beyond that, it seems that the attrition of Obama’s support between 2008 and 2012 was reasonably constant across sub-groups–and that Obama didn’t lose non-Hispanic white voters at markedly higher rate than he lost certain other demographics.