In response to my “request yesterday”:https://themonkeycage.org/2010/11/are_polls_more_likely_to_be_of.html for more information on bilingual polling, I learned the following.
First, according to “Dominique Arel, the Chair of Ukrainian Studies at the University of Ottawa”:http://www.ukrainianstudies.uottawa.ca/personnel/arel.html, no serious polling firm in Ukraine would ever employ anyone other than bilingual pollsters. The process works as follows:
bq. Regarding polling in Ukraine, the practice of the most established polling outfit (Kyiv International Institute of Sociology) is the following: the first question (in person, since serious polling can’t be done on the phone in Ukraine) is a linguistically neutral greeting, which could sound Ukrainian or Russian. The pollster then go with the language chosen by the respondent, if necessary, after specifically clarifying the matter. Serious organizations can’t do uni lingual polling in Ukraine, since it would necessarily offend a segment of the respondents–on each side.
Second, “Adam Berinsky”:http://web.mit.edu/berinsky/www/ and “Cara Wong”:http://www.carawong.org/ have an old, unpublished paper comparing bilingual with uni lingual interviews of Latinos. Here’s the abstract:
bq. It is common practice for academic and professional pollsters to conduct surveys in English only, and then weight up responses of the English-speaking Latinos to compensate for the missed Spanish-speaking households. While we recognize the practical reasons for interviewing in English only, there is little empirical evidence that this common practice is a “good” practice. In fact, it may distort our understanding of Latino public opinion. Our analysis indicates that Spanish and English respondents do differ in important. Nevertheless, the common polling practice of using English-only interviews and weighting for Latinos is an acceptable practice in most situations. While the top-line marginals obtained through such methods differ from the full population marginals in certain circumstances, the size of these differences ranges from small to nonexistent. However, if researchers are interested in describing the political beliefs of Latinos, bilingual interviewing is a necessity, not a luxury.
Third, Effram Perez has a 2009 _Journal of Politics_ article (“gated”:http://journals.cambridge.org/action/displayAbstract?fromPage=online&aid=6390168) on the topic as well. Here’s the abstract from his paper:
bq. The dramatic increase in the U.S. Latino population in recent decades has spurred an equally dramatic rise in bilingual survey instruments used by scholars to gauge the political attitudes of this growing ethnic group. A key assumption behind these instruments is that English-language items tap the same political constructs as their Spanish-language analogs. This paper reports evidence which suggests that bilingual survey items may not always be comparable across linguistic groups. Using a variety of public opinion polls, I develop and test a series of multi-group measurement models showing that–net of measurement error–English- and Spanish-language survey items are not functionally equivalent. The paper discusses the implications of these findings for the development of future bilingual surveys, both in the United States and beyond, as well as the use of extant surveys for applied analyses of Latino political attitudes.
To the extent that it is correct that the biggest mistakes in poll-based forecasts in the 2010 elections were made in states with “large Latino populations”:https://themonkeycage.org/2010/11/are_polls_more_likely_to_be_of.html, this is an issue that is probably worth keeping an eye on moving forward.