Abstract

Survey interviewers are usually required to follow standardized guidelines for conducting interviews. The introduction of computer-assissted interviewing facilitated the collection of interview audio recordings to assess whether interviewers indeed follow the guidelines. However, in most cases, audio recordings require respondent consent, and thus interviewers know whether an interview is recorded or not. In this paper, we assess whether interviewers change their behavior when interviews are recorded. Using automatically collected timestamp data and both longitudinal and cross-sectional analysis approaches to account for selection biases, we find that interviewers substantially change their behavior when recorded. Concerning substantive outcomes, we find no evidence for changes in average responses when interviews are recorded, but higher interviewer effects for recorded interviews.