One seemingly tedious task that I've come to enjoy is recoding the race of our respondents who say they're of 'some other race,' and then offer an answer. You never know what you're going to get
There are constants: Irish and Italian-Americans identifying as such. A handful of racists: "Aryan." Ridiculous answers: "Ginger," "a big fat gay mulatto!" Interesting detective work, where an interviewer types an obscure answer phonetically--"Chupic"--and you figure it out
('Chupic" appeared to be 'Cup'ik' an Alaska Native tribe, and the respondent was in the geographic part of western Alaska for it, which was fun)
One overarching takeaway from handling self-reported race this year, in a few contexts, is that this is a more challenging category for weighting than we might think. I'm not sure whether it's as easy to weight to census racial targets as I had assumed
In our South Carolina poll this AM, for instance, 26% of registered voters identify as 'black' when they filled out their voter registration form, with the question as shown here 
Same people, different answers in different settings:
Voter-file race: White 70, Black 26, Hispanic 2, Other 1
Self-report race on the poll: White 63, Black 24, Hispanic 5, Other 3, DK 5
Voter-file race: White 70, Black 26, Hispanic 2, Other 1
Self-report race on the poll: White 63, Black 24, Hispanic 5, Other 3, DK 5
You might wonder whether we have too few consistent white voters, and therefore it's too diverse/Dem. Yet it turns out that the folks listed as white on the voter file, but didn't call themselves white in the interview, are even more conservative than consistent white voters
And to be clear, this is after recording the "Aryan" and "Irish-American" folks as 'white.'
• • •
Missing some Tweet in this thread? You can try to
force a refresh
