Kristen Olson, Jolene Smyth, Jennifer Dykema, Allyson Holbrook, Frauke Kreuter, Brady T. West
The Past, Present, and Future of Research on Interviewer Effects

S. 22-33 in: Kristen Olson, Jolene Smyth, Jennifer Dykema, Allyson Holbrook, Frauke Kreuter, Brady T. West (Hrsg.): Interviewer Effects from a Total Survey Error Perspective. 2020. Boca Raton: CRC Press
[Chapman and Hall/CRC Statistics in Social and Behavioral Sciences]

In an increasing number of surveys, interviewers are tasked with collecting blood, saliva, and other biomeasures, and asking survey respondents for consent to link survey data to administrative records. Errors introduced by interviewers can take the form of bias or variance. Early research found that interviewers vary in how they administer survey questions and that their effects were similar to sample clusters in both face-to-face and telephone surveys. Given the nesting of respondents within interviewers, hierarchical or random effects models have long been used for the study of interviewer effects. Multilevel models are flexible and can be used to infer whether interviewer effects differ across subgroups of items, respondents, and interviewers. At the bare minimum, an anonymized interviewer ID variable on data files would allow analysts to estimate interviewer variance components. Additional data on interviewers, extending beyond simply demographics and experience, would facilitate understanding the mechanisms by which interviewers affect survey data.