Blog

In Florida, Are Polls Accurately Predicting the Hispanic Vote?

by Mahwish Khan on 10/29/2012 at 11:46am

Earlier today, ImpreMedia/Latino Decisions released its latest tracking poll of the Presidential race, finding that Obama leads Romney by a whopping 73% – 21% margin among Latino voters. Much of the focus is, of course, on the key battleground states, including Colorado, Florida, Nevada, and Virginia, where Latinos comprise key voting blocs.

As we’ve reported, a number of pundits have commented on what we’re calling the  “The Harry Reid Effect”:

The Harry Reid Effect happens when pre-election polls underestimate the Latino vote in such a way that seriously undercounts the actual support enjoyed by a candidate.  The methodological and sampling errors in flawed pre-election polls combine with flawed voter turnout models (that underestimate Latino voter registration and turnout operations) to paint an overall misleading picture.

Today, Nate Cohn at the New Republic takes a look at polling of the Hispanic vote in Florida. In addition to the issue of whether pollsters are underestimating the Latino vote, there’s a breakdown in Florida between Cubans and non-Cuban Hispanics, which may not always be reflected:

Florida polls show a tighter race than expected among Florida’s Hispanic voters. Rasmussen, Susquehanna, and Mason Dixon show Romney ahead by 3 or 4 points, while NBC/WSJ/Marist, PPP, and Fox News all show Obama with less than a 5 point advantage. And finally, a well-publicized FIU/Miami Herald/El Nuevo Herald Poll showed Obama ahead by just 6 points in a poll of Florida Hispanics. But two polls show Obama maintaining support near ’08 levels among Florida’s Hispanics. CNN shows Obama up 74-24 among non-white voters, the same as his 74-25 advantage from 2008. SurveyUSA shows Obama leading among Hispanics by 17 points. It’s probably not a coincidence that the latter two surveys show Obama leading and Romney up by just one.

For Obama to fall off so far among Florida Hispanics, non-Cuban Hispanics would need to stay home and Romney would need a strong showing among Cuban voters. Both halves seem possible, but the voter registration numbers suggest that demographic changes are sufficient to swamp declining turnout rates. So generating a ten point swing among Florida’s Hispanic population through a decline in support among Florida’s Cuban voters would probably require a 20 point swing among Cubans, given that Cubans will probably represent something slightly less than half of Florida’s Hispanic voters, and perhaps much less than half (they’re about 30 percent of registered Florida Hispanics, but they turn out at higher rates).

The issue might be that polls don’t weight Hispanic voters by ancestry, so if 60 percent of respondents are Cuban, polls won’t take measures to ensure that a larger share are Mexican-American or Puerto Rican. This introduces another layer of error into the already difficult task of polling Hispanic voters, and it might be influencing the numbers. Two data points are consistent with this possibility. A Latino Decisions survey found Obama ahead by 30 points in early October, and their surveys are weighted for ancestry. Similarly, SurveyUSA, the only statewide, full-population survey to find Obama leading by a double digit margin among Hispanics, disaggregated and presumably weighted by both Cuban and non-Cuban Hispanics and found Obama ahead by 40 points among non-Cuban Hispanics but trailing by 8 among Cubans, with Cubans representing 48 percent of Hispanics. Those numbers sound about right and that might have something to do with why SurveyUSA is the only recent poll showing Obama ahead in the Sunshine State.

It’s great to see reporters, like Nate Cohn, dig deeper into the Latino polling. We’re seeing more of this kind of discussion in 2012, including an analysis from Nate Silver at the NYT’s FiveThirtyEight. We’ll need this same kind of digging into exit polls on Election Day.

Previous post:

Next post: