The conservative National Review declared, “Republicans made historic gains across the country on [election day], including significant progress with minority voters.”  The Houston Chronicle and Fox News Latino similarly pronounced Latino voters having a large voting preference for Republicans in Texas, Kansas, and Georgia.

But these claims are based on exit polls conducted by market research firm Edison Research – polls judged as “methodologically indefensible” by a group of political scientists at Latino Decisions, a firm specializing in Latino public opinion research.  A principal at Latino Decisions and also a professor of political science at Stanford University, Gary Segura wrote in an e-mail, “Those numbers are methodologically indefensible and they know it.  NEP [Edison Research’s National Election Pool] has to tell us how many interviews they conducted, where they conducted them, and how many were in Spanish.  Without that information, these claims should not be taken seriously.” In the interest of full disclosure, I have also worked with Latino Decisions in the past.

Edison Research states its exit polls are designed to “group together precincts with similar vote characteristics” to account for partisan division in each precinct based on past elections.   Justin Gross, the Chief Statistician at Latino Decisions and a political scientist at the University of North Carolina, wrote in an e-mail that Edison Research’s methodology is focused more on state predictions rather than group-specific understandings: “…the notion of being representative is always with respect to specific traits, and for Edison, that surely includes race and ethnicity, but not attributes reflecting diversity among Latinos, such as language preference or nation of ancestry.”

Edison Research officials responded in a call with more details of their methods, namely a list of four precincts covered in their exit polls in Texas that contained a majority of Latinos: precincts in El Paso (2), Harris (1) and Travis (1) counties. (Other Latinos in the NEP in Texas were from a sample frame of 21 precincts throughout the state.)  The question, then, is whether or not the interviews gathered from this selection of precincts would be representative of Latinos across Texas.

A large number of Latino voters speak Spanish, an important characteristic that Segura argues should be represented in any poll making predictions about Latino voters.  When asked over the phone for details about the number of surveys conducted in Spanish, Joe Lenski, executive vice president at Edison Research, said, “I’ll e-mail you those numbers as soon as I get them.”   Those numbers have not yet been sent.

For its part, Edison Research argues Latino Decisions is concentrating its phone interviews in high-density locations, which means it reaches Latinos who are more likely to vote Democrat.  Lenski says about Latino Decisions, “They’ve been barking about this for 20 years.  It’s a difference in methodology.

This “difference in methodology” came to a head in the 2004 presidential election, when the exit polls showed 44 percent of Latinos voted for George W. Bush. Academics disputed this and Edison Research along with Mitofsky International stood firm on their numbers, even responding with an article published in a widely read academic journal. In it, Warren Mitfosky writes:

“I think the article by David Leal et al. is wrong in its conclusion. The Los Angeles Times and NEP exit polls were a more appropriate method for estimating the Hispanic vote than the pre-election surveys the authors used as the basis of their arguments. Most of these surveys were taken months before the election. The methods for sampling Hispanic voters are not given.”

The central debate is on whose opinions are captured. At a basic level, the respondents participating in polls must be representative of the population analysts are making predictions about. On whether Edison Research has a representative sample of Latinos in Texas, Segura says,  “In a state as large as Texas, interviewing Latinos in only 25 locations, and effectively none in the Rio Grande Valley, introduces a bias that is just insurmountable.”

Tedious debates over research methodology seldom make headlines, but the consequences for these differences can be profound for politics.  The GOP can take comfort in numbers suggesting it’s made inroads with Latino voters and be lulled into thinking it won’t suffer a backlash in future elections from its hard stance on immigration. Mismeasurement of Latino electoral support could also discourage Democrats from investing in Latino voters in conservative states with a growing population of Latino voters, like in Texas and Arizona.

Getting the Latino vote right will matter a great deal in 2016, when the electorate will be more diverse across the states and across the nation. As Latinos turn out to vote in greater numbers, basing future predictions on past performance risks ignoring the details about Latino voters.  Pollsters demonstrate a nuanced understanding of white voters, using such terms as “soccer moms” and “NASCAR dads,” but the same appreciation for similar nuances among Latino voters still hasn’t emerged.

The race to understanding these differences has its own self-affirming impact so long as Republicans are winning elections, even if for the wrong reasons.  The short-term toll for these inaccuracies, if Latino Decisions is accurate, will be paid for by Latinos in the form of policies that are incongruent or hostile to their needs and their experiences. In the long term, however, it may be the Republican Party that pays the cost for these differences in methodology.


Stephen A. Nuño is associate professor of political science at Northern Arizona University, where he studies political behavior, race and ethnic politics, Latino politics, mobilization, and partisanship. Follow him on Twitter at @stephenanuno.