Race of Interviewer Effects and Interviewer Clustering

Traditional approaches to understanding race of interviewer (ROI) effects have focused more on respondent error than interviewer and survey design error. Respondents are hypothesized to alter their responses to certain items in an effort to maintain a favorable impression for the interviewer. Yet, this framework assumes respondents, interviewers, and questions all work in concert to produce the effect. The variability among these components of the survey system actually makes producing a statistically significant interviewer quite difficult. We argue that it is those interviewers who are assigned higher workloads, and therefore interact with more respondents, who are exacerbating the ROI effects to the point of statistical significance. We analyze individual and aggregate pre-election data from the 1984 National Black Election Study (NBES), finding support for our hypotheses. Feeling thermometer scores that were initially influenced by the interviewer's race, were subsequently reduced to non-significance when interviewer workload was considered. Our findings suggest the often elusive ROI effect is potentially more related to survey design features than psychological processes related to impression management