Paradata related to reasons for refusals in face-to-face surveys can be useful for research on refusers and on refusal rates. This article addresses the measurement quality of reasons for refusals. Multilevel multinomial models were used to examine the effects of interviewers on the collection of reasons for refusals in a survey on the German population, the General German Social Survey (ALLBUS 2008). In this survey, reasons for refusals were collected by interviewers, predominantly in an open-ended answer format. We categorized reasons for refusals - our dependent variable - through content analysis. The results of the subsequent multilevel analysis show that interviewers' gender, age, and workload are related to the collection of some reasons for refusals. At the same time, we obtained high unexplained interviewer variance, which indicates that interviewers tend to report certain reasons for refusals. Further research should investigate how to reduce the effects of interviewers on collection of paradata.