Recently, there was a story claiming Republican candidates were winning among Latino voters in Georgia. The entirety of this story was based on a small portion of a SurveyUSA/11Alive poll that interviewed a grand total of 36—yup, three dozen—self-identified Hispanic voters.
So was the poll right? Should we be worried?
Not a chance. On the contrary, this is a good example of why you should never trust the crosstabs of a minority demographic, especially in robopolls. (And that goes for all types of demographic minorities, for example age 18-29.) You simply cannot get a good measure of opinion with only a handful of respondents.
To see an example of Latino voter data from ordinary polls failing miserably, here's the data from the crosstabs of October 2012 Colorado polls of the presidential race, compared to the exit poll, and the two Latino Decisions polls (which poll only Latino voters and are carefully designed to accurately measure the opinions of the Latino electorate).
Our best measurements—the exit poll and Latino Decisions polls of just Latino voters—show Obama receiving between 74-87 percent of the Latino vote. But if you had looked at ordinary polls, they would have told you Obama was set to get only about 48-66 percent—a gross underestimation.
None of the ordinary polls was designed specifically to measure the Latino electorate, and they clearly failed. This is typical for most states. But the Latino Decisions polls were close to the exit poll (and arguably, the election eve Latino Decisions polls are more accurate than the exit polls).
These are the polls you are looking for
Fortunately, Pew and Latino Decisions/America's Voice have recently published national surveys of Latino voters. Latino Decisions has also released polls of Colorado, Florida, and North Carolina, done for the National Council for La Raza Action Fund. And there's some good news, and some bad news.
The good news comes in the topline numbers for Senate and governor: Latino support for the Democratic candidate is on course to equal 2012 levels.
The bad news is that Latino voters are not as happy with Democrats in general as they have been in the past, most likely due to a lack of action on immigration issues.
Help elect more and better Democrats this November! Please give $3 to Daily Kos' endorsed candidates and strike a blow against Republicans.
Defeat Mitch McConnell in just two hours. Sign up to make GOTV calls to Democrats.
Please read below the fold for more on Latino voters and polling.
Topline numbers
In Florida, 73 percent of Latino voters say they are certain to vote. They back former Gov. Charlie Crist 53-29 over Gov. Rick Scott. This is a huge reversal from 2010, when Scott won 51 percent of the Latino vote. In September 2012, 52 percent of Latino voters said they backed the Democratic House candidate; by election eve, that number had grown to 58 percent. Current Latino support for Crist is at 2012 levels.
In Colorado, 80 percent of Latino voters say they are certain to vote. Sixty-six percent say they support Senator Mark Udall compared to 17 percent for Cory Gardner. This is very close to the 68 percent that said they supported the Democratic House candidate in 2012 in the October poll. By election eve, that number had grown to 88 percent. Again, we see Latino support for Democrats at 2012 levels.
In North Carolina, 69 percent of Latino voters say they are certain to vote and they support Senator Kay Hagan 52 percent to 22 percent. On election eve in 2012, 70 percent supported the Democrat for Congress.
In the Latino Decisions nationwide poll, 71 percent of Latino voters say they are certain to vote and 59 percent support the Democratic House candidate, compared to 25 percent for the Republican. This is down a little from 2012, and more comparable to 2010.
In the Pew nationwide poll, 53 percent of Latino voters say they are "absolutely certain" to vote, just slightly more than 2010. Voters support Democrats over Republicans 57-27 for the Congressional ballot, a little less than 2010 numbers from Pew.
Warning signs
Latino voters may still break strongly for Democrats, but there's many clear warning signs in the national poll. For instance, in the Latino Decisions poll, only 28 percent say they are voting because they want to support the Democratic candidate; this is down from 38 percent in 2012. Fifty-one percent disapprove of how Obama is handling immigration, and 53 percent disapprove of how Democrats in Congress are handling immigration.
This is of no small importance, as 51 percent of respondents listed immigration policy as one of their top two most-important issues in an open-ended question. In another question, 66 percent said immigration was one of the most important issues or the most important issue in their voting decisions.
The Pew poll shows a slight decrease since 2012 in the Democrats' partisan identification advantage among Latino voters. Seventy-three percent say immigration is an issue that is extremely or very important to them personally, although health care, jobs/economy, and Education come out ahead. Democrats still hold a 23-point advantage over Republicans in dealing with immigration, but 34 percent blame Democrats for not passing reform yet and 35 percent are angry or disappointed about Obama's decision to delay executive action relating to immigration. Finally, 55 percent of Latino voters disapprove of Obama's handling of deportations.
The lack of progress on the immigration issues may be contributing to the feeling that Democrats aren't doing a good job reaching out to Latino voters. In the state polls, only 41 percent (CO), 40 percent (FL), and 31 percent (NC) say Democrats are doing a good job reaching out to Latino voters. That's still much better than Republicans, but not indicative of a strongly supportive relationship.
Wait, but why can't you trust the crosstabs?
The short answer is that small samples, English-only interviews, and a poor understanding of demographics can combine to remove any meaningful information from the subset of voters identified as Hispanic or Latino in many polls. The folks at Latino Decisions have written extensively on this topic in recent days. In addition, in robopolls at least, many of the respondents may be incorrectly identified as Latino voters when they really are not. This conclusion arises from analysis of the 2012 DailyKos/SEIU/PPP polling data. All this combines to render the crosstabs for Latino voters in most polls essentially worthless, unless the poll has been carefully designed with the Latino electorate in mind.
Help elect more and better Democrats this November! Please give $3 to Daily Kos' endorsed candidates and strike a blow against Republicans.
Defeat Mitch McConnell in just two hours. Sign up to make GOTV calls to Democrats.
