Interviewer Behavior and Survey Data Quality: The Case of Social Network Data

• Activity Number: 232
• Type: Topic Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Bernadette Huyer-May*+ and Josef Brüderl and Claudia Schmiedeberg
• Companies: and University of Munich and University of Munich
• Keywords: Interviewer effects ; interviewer deviant behavior ; filter questions ; jackknife ; measurement error ; network data

Abstract

If interviewers are paid per interview completed, they have strong incentives to shorten the interview for instance by answering filter questions so as to skip lengthy follow-up questions. If many interviewers show such "rational" behavior data quality will be damaged seriously. We provide evidence from a large german panel, that such rational interviewer behavior is indeed widespread. We find that ego-centered networks generated in waves 2 and 4 of the Family Panel are surprisingly small. We also find large interviewer effects which cannot be explained by interviewer or respondent characteristics. We try to identify those interviewers who are particularly responsible for the large interviewer effects by using the jack-knife. We identify three groups of interviewers: Those who elicit very small networks ("fraudulent" interviewers), those who generate particularly large networks ("diligent" inter-viewers), and the rest. Further we show that only the diligent interviewers provide networks of a reasonable size. In addition, substantial results obtained with network data from normal/fraudulent interviewers differ from those obtained with data from the diligent interviewers.