Why DHM Will Never Have a Robot Call You

Share this post:
By Paul Gronke, DHM Research

Like everyone, we’re inhaling every snippet of news these days, trying to get a good read about the upcoming election. That’s why this story grabbed our eye: “Hispanic Voters Buck Assumptions: Back GA GOP Candidates.

If true, this would be quite a story. Latino voters are a rapidly growing segment of the electorate.  Latinos tend to be religious and socially conservative but liberal on issues of immigration and other economic issues.  Latino voters were key to both of Obama’s victories, delivering nearly 75% of their votes for the Democratic ticket—15% more than voted for John Kerry in 2004.  Cubans, the one reliably Republican group among Latinos, now show only a tiny Republican advantage over Democrats (47% - 44%).  (Look here to see how Oregon’s Latino electorate compares to Latinos nationwide.)

What’s up in the Peach State?  It turns out, nothing much at all, other than bad survey methodology. SurveyUSA, the firm that conducted the survey, relies on phone calls using the “recorded voice of a professional announcer.”  In other words, robo-calls.

What are the problems with robo-calls?  Robo-calls have a “Republican house effect” as high as four percent. And the surveys are conducted only in English, excluding any respondent who wants to take the poll in Spanish.  The result of all this is that the news story turns out to be based on only 38 Latino respondents!  In short, the story is bunk.

This is why, at DHM, we’ll never have a robot call you.

DHM Research relies on three kinds of survey methods—telephone interviews, internet surveys, and focus groups—and three kinds of samples—random digit dialing (for telephone), randomly selected online surveys, and online panels.  We pay close attention to cutting edge academic research on the use of online panels and internet surveys in particular, to make sure we avoid any kind of “house bias.”

In fact, we’re pleased to report that Nate Silver recently found DHM's party bias to be zero point zero.