Extract

Interviewer Behavior and the Quality of Social Network Data1 Josef Brüderl, Bernadette Huyer-May, Claudia Schmiedeberg Abstract Interviewer effects are a typical – although often neglected – phenomenon of social network data collected in personal interviews. We analyze the ego-centered network data provided by the German Family Panel and find large interviewer effects which cannot be explained by interviewer or respondent characteristics. These interviewer effects are caused to a large degree by two groups of interviewers, i.e. those who elicit less network persons than the average (“fraudulent” interviewers) and those who generate particularly large networks (“diligent” interviewers). We suggest a method to identify these groups of interviewers. Introduction In recent years, social network analysis has gained popularity (for an overview see e.g. Wassermann and Faust 2009). The majority of social network studies have relied on data gathered by standardized personal interviews (Marsden 1990, Matzat and Snijders 2010). But caution is advised as social network data are vulnerable to “noise” (Fischer 2009) as well as to interviewer and design ef- fects. These effects may result from deviating interviewer behavior. In contrast to faked interviews, which most of the other chapters of this volume (e.g. Bredl et al. 2013, Menold et al. 2013) deal with, in our case the interviewers’ devia- tions are limited to a small part of the interview whereas other parts of the data should not be affected. Nevertheless, according to AAPOR (2003) this can be seen as a (partial) falsification of interviews. In the research tradition on falsifi- cations generated...