Anti-Immigrant Sentiment and Occupational Context: An Examination of Multilevel Model Estimates When Samples Are Small

Many who study anti-immigrant sentiment attribute negative attitudes among the native population to objective economic threats that immigrants may pose. In multilevel studies, researchers focus almost exclusively on geographic regions, such as metropolitan areas or countries, as contexts within which to examine the consequences of objective economic threats. Although geographic regions are relevant, it is important to measure competition in other contextual units, such as occupations. Methodological challenges, however, have inhibited the measurement of economic competition and other important concepts in alternative contexts. Small sample sizes within occupations, for example, raise questions about statistical power and estimation. In this paper, the author uses data from the 2004 General Social Survey (GSS) to examine the consequences of small occupation-specifi c sample sizes for multilevel models predicting the perceived threat of immigrants in the US. The author examines estimates using different groupings within the International Standard Classifi cation of Occupations (ISCO) scheme: 1) 390 detailed occupations, 2) 116 minor groups, 3) 28 sub-major groups and 4) 9 major groups. Results demonstrate that estimates based on a larger number of occupations (i.e., 390 or 116) are generally adequate despite the small occupation-specifi c sample sizes. Moreover, pooling the data substantially reduces the between-occupation variance, which may lead researchers to conclude that occupations are irrelevant.