Nate Silver has an interesting post that highlights some of the cases his model got wrong. In particular, his “poll of polls” approach missed Nevada and Colorado, and underestimated significantly margins of victory in California. This suggests there might have been something systematically wrong with the polling in those three states (as opposed to more general concerns about polling problems that should have been present more widely, like the cell-phone only households).
Matt Barreto, a political science professor at the University of Washington and one of the authors of the excellent Latino Decisions Blog, wrote in to Silver and offered the following explanation. In Silver’s words:
His [Barreto’s] firm, which conducts interviews in both English and Spanish, had found that Latino voters—somewhat against the conventional wisdom—were relatively engaged by this election and for the most part were going to vote Democratic. Mr. Barreto also found that Latino voters who prefer to speak Spanish—about 40 percent of Latino voters in California meet this description, he told me—are particularly likely to vote Democratic. Pollsters who don’t conduct bilingual interviewing at all, or who make it cumbersome for the respondent to take the poll in Spanish, may be missing these voters.
Does anyone out there have any research on bilingual polling? Maybe someone who has worked on polling in Ukraine or Georgia? If so, please respond below in the comments section, or contact me directly if you’d like to do a guest post. A quick Google Scholar search for “bilingual polling” revealed only one relevant article, a 2001 Hispanic Journal of Behavioral Sciences article by Kevin A. Hill and Dario V. Moreno. Here’s the abstract of their piece:
This article argues that conducting public opinion surveys in Spanish as well as English is crucial to the study of the modern Latino electorate. Unfortunately, the conventional wisdom is to survey only in English because, so the argument goes, the validity and reliability problems raised by bilingual polling and translation do not make it worthwhile to conduct surveys in two languages. The authors challenge this assertion with evidence from six political surveys in Miami-Dade County, Florida, that were conducted in both English and Spanish. It is found that, had the conventional wisdom been followed and the polls been conducted in English only, results would have been profoundly inaccurate and invalid. The authors further take advantage of bilingual survey research methodology and assess the level of difference between the survey responses of English-speaking and Spanish-speaking Latino voters, comparing the former to non-Hispanic White voters as well. It was found that, on average, English-speaking Hispanic voters gave sets of responses to different survey questions that were roughly equidistant between those of non-Hispanic Whites and Spanish-dominant Hispanics. The importance of these findings, not only for the survey research methods literature but also for assimilationist models of ethnicity, is assessed.
Full text is available here, gated.