10 Caveats about Language Data from Census 2000
To understand Census data on language trends and their implications, it's essential to understand how the information was gathered:
1. These numbers are based entirely on "self-reports" – respondents' own subjective assessments of their language usage and proficiency. Follow-up studies in this area have shown that self-reports tend to be unreliable.
2. Census questions about language are abbreviated, ambiguous, and – for many people – confusing (see below). No explanatory notes or objective criteria are provided to help respondents in answering them. So they yield less accurate information than more careful surveys do, according to critics (e.g., Veltman, 1988).
3. Perhaps the best to be said for these questions is that the Census has asked them consistently since 1980, rather than changing their focus every ten years, as in the past. So it is possible to plot trends for the past two decades – 1980-1990 and 1990-2000 – without having to juggle apples and oranges.
4. Census data about language are estimates based on the "long form" questionnaire sent to a sample of about 20 million (one in six) U.S. households in 2000. Detailed results from this survey are scheduled to be released by early 2003.
5. Preliminary results were announced in August 2001 from the Census
2000 Supplementary Survey of about 700,000 households. These figures
offer a preview of the full Census findings but are no doubt even less
a. Does this person speak a language other than7. In pondering question (a), "Does this person speak a language other than English at home?" a respondent might reasonably ask:
8. The net effect of these ambiguities is almost certainly to overstate
the U.S. language minority population as it is commonly understood: persons
who maintain fluency in a native language other than English. A substantial
number of respondents from native-English-speaking backgrounds seem to
be reporting the use of other languages. In the 1980 Census, for example,
28 percent of those who said they spoke Spanish at home also reported non-Hispanic
ethnicity – hardly a credible finding, considering
how few Anglo-Americans are fluent bilinguals.
10. On the other hand, the Census has acknowledged that it substantially undercounts immigrants and Native Americans. This problem obviously tends to depress its estimate of linguistic diversity. In 1990, for example, it failed to record speakers of more than 50 Native North American languages that were still spoken at the time.
O'Malley, J.M. 1981. Children's English and Services Study: Language Minority Children with Limited English Proficiency in the United States. Rosslyn, VA: National Clearinghouse for Bilingual Education.
Copyright © 2001 by James Crawford. Permission is hereby granted to reproduce this page for free, noncommercial distribution, provided that credit is given and this notice is included. Requests for permission to reproduce in any other form should be emailed to email@example.com.But before writing, please read my permissions FAQ.