The Problem With Polling Latinos

PHOTO: President Barack Obama waves to supporters as members of the Mexican Rock band Mana walk off stage during campaign event.Pablo Martinez Monsivais/AP Photo
President Barack Obama waves to supporters as members of the Mexican Rock band Mana walk off stage during a campaign event in Desert Pines High School, Sunday, Sept. 30, 2012 in Las Vegas. Mana performed at Obama?s campaign stop in Las Vegas, continuing the campaign?s effort to rally Hispanic voters who stand to sway the election in the swing state of Nevada.

As pundits and politicos make proclamations about how Latinos will cast their votes, pollsters warn that Latino voters are still the hardest of all to measure due to barriers unique to the population.

Most voter polls throughout the election season have found that considerably more Latinos will be casting ballots for Obama than Romney by roughly 2 to 1.

Before Democrats declare certain victory though, many pollsters warn that getting an accurate handle on Latino voters is a harder task than with any other voting group because Hispanics in particular tend to be young, multi-lingual, and mobile -- and therefore more expensive (and difficult) to survey.

"There's always a trade off in polls between cost and perfect coverage of the population," said Trevor Tompson, the Principal Research Scientist and Director, Associated Press-NORC Center for Public Affairs Research. "You never have an unlimited budget."

The high price tag for polling Latinos has limited the number of trusted polling institutions that consistently survey the community, with some of the largest studies costing upwards of a quarter of a million dollars. Pew Hispanic Center, Bendixen & Amandi, and Latino Decisions (which Univision has partnered with to conduct polls) have emerged as some of the most prominent in the field, although each have different philosophies about where corners can be cut to save money (and how many of such cuts can be made and still be able to consider a measure trustworthy.)

So why is polling Latinos in particular so expensive?

For one, hiring bilingual interviewers is a must, says Pew Hispanic Center's Associate Director Mark Hugo Lopez who noted that over half of Latinos polled at his institution choose to do their survey in Spanish. To not allow for Spanish respondents, or to require a "call back by a Spanish language interviewer drastically cripples the survey's credibility, even though these options might be cheaper than hiring bilinguals.

Another issue is cell phone usage. Because Latinos skew younger and are less likely to be anchored in one home for a long time, it is important for polling agencies to have access to cell phone number lists, in addition to landline phones -- a luxury which also adds to the cost of the survey.

Sampling Latinos as a small fraction of a national survey can also often problematic, pollsters say. Fernand Amandi from Bendixen & Amandi (a firm that conducts polling for President Obama) notes that translating long-winded English surveys into Spanish only makes them longer, and Hispanics have less patience when it comes to answering lengthy surveys, and will cut the survey short more often.

"There's only so much tolerance in Latino culture for these 85 question Odysseys," Amandi said. "Especially when you consider that it takes 15 to 13 percent longer to do the same interview in Spanish than in English."

Often times, national polls also fail to survey enough Latinos or break out the demographics by important indicators such as state, age, foriegn born vs. native born, and country of origin. A recent poll from Quinnipiac took a sample of just 143 Latinos as part of a greater national poll. Only 14 percent of these interviews conducted in Spanish, a number the Latino Decisions pollsters call "an unacceptably low number."

"National polls are not designed to get an accurate geographic representation of Latinos because they draw a nationally proportionate sample of all Americans, and pick up Latino respondents wherever they surface. The Latino population maintains substantially different patterns of residence than the national population and as such, the Latino sample is rarely representative of the overall Latino population," the Latino Decisions pollsters wrote on their blog.

Polling Latinos alone is therefore preferable, but this of course requires for institutions to shell out more cash. In order for Pew Hispanic Center to achieve a sample size of 1,700 Hispanics by calling random numbers (called a random digit dial technique), Lopez estimates that approximately 240,000 phone calls have to be placed by their bilingual interviewers. Pew is known to be one of the best Latino pollsters in part because of their large endowment that comes from the Pew Charitable Trust.

Still, Pew Hispanic often avoids week-by-week voter polling in part because their methods are time-consuming, making it difficult to release results that are up to date.

Some institutions choose to find interviewees by calling those on registered voter lists that have Hispanic-sounding surnames, in order to shorten survey times and save money. But this method also has its own pitfalls.

"So many Latinos are first-time voters. They don't appear on these lists until just before the election," said the Associated Press-NORC's Thomson.

But of all the unique hurdles to polling the Latino community, the hardest and most important is knowing who is actually going to show up on November 6th.

"With all of the errors that are caused with short cuts in polling, the voter turnout estimates have such a huge margin of error around them, with 10 percentage point swings that wouldn't be detected," Tompson said.

While Latinos will certainly swing Democrat, no one knows how many will truly vote when the time comes, particularly in swing states.

"We do have to remember that certainly the polling methods that we have now are good enough to tell us that Latinos are mostly Democrats at the moment," Thompson said

"What the polls can't tell us now are all the nuances, and the divisions between regions and ethnic groups," he added. "And especially the question of voter turnout -- the polling methods we're using now don't do that well at all."