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 looks at the relationship between Republican opposition to immigration and the party's troubles with Latino voters. For past posts in the series, head here.

Pundits left and right have embraced the notion that the Republican Party has a strong political interest in passing comprehensive immigration reform.  As the argument goes, the GOP lost a good shot at the presidency in 2012 because of a pro-Obama shift among Latino voters, which itself was a reaction to the GOP’s increasingly anti-immigration stance.

There is circumstantial evidence backing that view.  According to exit polls, the share of Latino voters supporting the GOP candidate dropped from approximately 40 percent in 2004 to 27 percent in 2012.  An election eve phone survey by Latino Decisions in 2012 put Mitt Romney’s share even lower, at 23 percent.

Still, others doubt that a visible repositioning by the GOP would lead to a sizable change in Latino voting patterns.  That’s the view of Matt Continetti, who argued in The Weekly Standard that “[i]llegal immigration is not the reason Hispanic voters support the Democratic party. Hispanic voters support the Democratic party because they tend to agree with its domestic policy agenda of redistributing money to the middle class and needy.” The question then becomes, why have exit polls recorded such a shift in Latino support over the past few presidential elections?  If the Latino electorate is shifting away from the GOP primarily because of changes in its composition, and not changes in the preferences of individual Latino voters, Continetti could be onto something.

So what does the evidence say about how Latinos have responded to changes in the GOP stance on immigration? Here, I look at two examples, both of which tell a similar story.  Latino voters do turn anti-Republican in reaction to Republicans who are perceived to turn anti-immigration.

The first example comes from Maricopa County Sheriff Joe Arpaio.  In 2004, the Phoenix-area official was reelected with 57 percent of the vote.  Beginning in 2005, he gained national attention for a crackdown on unauthorized immigration that included sweeps of day laborer centers and well-known detention facilities.  In 2008, he was reelected with 55 percent of the vote.  But what about his Latino constituents?

To get a sense of their voting patterns, I merged precinct-level election returns in those two elections with Census data on local demographics.  Taking into account the GOP presidential candidate’s performance in each precinct and other demographic measures, I then estimated the relationship between each precinct’s percent Latino and its support for Sheriff Arpaio.

Even before the crackdown, Joe Arpaio’s support declined as precincts grew more Latino.  Moving from a precinct that is 25 percent Latino to one that is 75 percent Latino, we would expect him to lose about 5 percentage points in 2004 once we adjust for presidential partisanship and other demographics.  After the crackdown, that same figure grew to 8 percentage points.  America’s Toughest Sheriff does seem to have paid a penalty in Latino precincts in 2008, even in a local race that wasn’t very competitive—and even though he had relatively little Latino support to begin with.

Still, there are limitations to precinct-level studies.  Chief among them, we don’t know which individuals’ choices are changing, or whether the changes are instead driven by who is turning out to vote. So let’s turn to panel data, data which surveys the same individuals at multiple points in time.  In October 2012, researchers from the University of Pennsylvania and Georgetown University fielded one wave of an ongoing panel survey that had begun in 2008. Using this Institute for the Study of Citizens and Politics data, we can investigate how the same voters’ choices changed from 2008 to 2012.

I focus on respondents who were McCain supporters just prior to the 2008 election, and trace their presidential preferences to October 2012, during the height of the presidential campaign.  The figure below shows which demographic attributes predict heightened or diminished support for Governor Romney.  Arrows to the right indicate groups of McCain supporters who were more likely to remain in the GOP camp. We see that older McCain voters were especially likely to stick with Romney, and that swing state residents were less likely to do so.

Still, the question at hand is whether the Latino respondents became less likely to support the GOP ticket in 2012, after Arizona’s SB1070 and Romney’s advocacy of “self-deportation.”  As it turns out, Latino McCain supporters were more likely to leave the GOP camp than any other demographic group analyzed here.  McCain supporters who were not Latino stuck with Romney 84 percent of the time, while the senator’s Latino backers only stayed with Romney 70 percent of the time.

There are caveats to that conclusion.  The data include 80 Latinos who backed McCain, and so the estimates carry some uncertainty: the effect of being Latino is actually positive 5 times out of 100.  But the estimated effect is sizable.  And it suggests that it’s not just that the composition of the Latino electorate has been changing.  Individual Latino voters have moved away from the GOP in recent years.

Now, was it the GOP’s stance on immigration that influenced Latino voters?  The panel isn’t conclusive on that point. We know only that Latino voters left the GOP, not why. But considered alongside the Arpaio results, the different voting trends among non-Hispanic whites, and polls showing the importance of immigration to Latino voters, it’s certainly a leading explanation.  And what the panel does show is that changes in Latino voting are the product of changing minds as well as changing demographics.  To some extent, to say that Latinos voters are lost to the GOP is to ignore recent electoral history.