Looking at Monday’s polls, I noticed that in Colorado SurveyUSA was releasing two separate statewide polls for two separate clients the very same day. I thought “Huh. Wonder what’s up with that?” and promptly forgot about it.
But turns out, as The Upshot‘s Nate Cohn extensively discussed today, that SUSA not only did simultaneous polls for two clients, but used two different methodologies, which creates a rare and fascinating opportunity to compare their accuracy–on November 4, at least.
But interesting as the comparison of random-dialing versus registration-list-based polling undoubtedly is (and the related issue of how your reach cellphone users), something else jumped out at me from the interview Cohn conducted with SUSA’s Jay Leve, who in response to a question about Cory Gardner surprising strength among Latinos said this:
I don’t know that I’ve ever seen a good sample of Hispanics. It’s a real challenge. It’s a tough population to poll.
There’s a decision at the outset: Are you in a position to offer your clients bilingual interviews, and are they able to pay for polling both in Spanish and in English? When we give our clients the bilingual option, many decline for cost reasons. When they accept it, like Nevada in 2012, where all of our polling was bilingual, I don’t know that we got one micron closer to the Latino population than had we polled in English only. Our data was criticized at the time for being too Republican, for being at odds with common sense. I get that criticism; I understand it. And the Hispanic data that you’re looking at in Colorado, that shows a Republican ahead among Hispanics, is also at odds with common sense. So I can’t defend it except that we give people the opportunity to self-identify as Hispanic, and we record it.
We have been accused in the past as having blacks who are not “black enough.” I get that criticism. Our black respondents, instead of being 90-10 Democratic, are sometimes 67-33. Do I think it turns out that way on way on Election Day? No, I think we’re too Republican on black voters, just as we are sometimes too Republican on Hispanic voters. This is not unique to SurveyUSA.
Are there people who specialize in Latino polling who conduct elaborate studies and then in turn prove, to their satisfaction and probably mine, that the Latino population is overwhelmingly Democratic? Yes. Is there something that we can do better? I’m sure that there is. At the moment, though, it is what it is. It’s what the respondents tell us when we give them a chance to identify as Hispanic and we ask them for whom they’ll vote.
That will be another thing to pay attention to when the dust settles on November 5. Sounds to me like a Udall “upset” in Colorado (and I guess that’s what it would be unless Gardner’s recent polling surge abates) would not be that surprising.