Gender of Interviewer Effects in a Multi-topic Centralized CATI Panel Survey

Abstract

This paper is motivated by two recent articles which show that numerous studies which analyzed gender of interviewer effects did not take interviewer nonresponse selection effects into account. For example, interviewers may be more successful at recruiting respondents with characteristics similar to themselves and who give answers that are similar to their own, and this may result in spurious gender of interviewer effects. Our research is novel because it uses data from a large panel survey in which the same respondent is asked the same questions repeatedly by interviewers of random genders using the centralized telephone mode. We use the panel design to show the importance of checking for all relevant variables in models where selection may cause bias. To this end, we use respondent fixed effects models as a reference to yield unbiased coefficients.

We find gender of interviewer effects that are in line with social desirability theory on gender issues such as female discrimination. However, not all gender-related questions are affected by gender of interviewer effects and, in addition, we do not find any effects on political and (factual) household task related questions. In line with the notion of social distance, there is a higher likelihood that answers respondents are less comfortable with are given to interviewers of the same gender regarding (sensitive) health questions.