Abstract

High quality surveys at the IAB are conducted through interviewer ad-ministred surveys. The involvement of interviewers into the data collec-tion process has adverse effects on variance estimation. Data are clus-tered within interviewer and subsequently variances are inflated reduc-ing the efficiency. As a consequence costs for surveys are inflated. Design effects are well known in face-to-face surveys and Schnell & Kreuter (2005) demonstrated for German face-to-face surveys that in-terviewers contribute substantially to such design effects. Up to now interviewer induced measurement error is seen as the key factor contri-buting to this effect. However, hypotheses have been raised that the effects are due to selective participation in the survey rather than a broken interaction between interviewer and respondents. Thus a func-tion of nonresponse bias rather than measurmeent error. So far this hypothesis has only been tested for one telephone survey (West & Olson 2010) with a very specific set of administrative data and an uncomfortable amount of working assumptions. The IAB is in a unique position to test this hypothesis with its combined survey and administrative data sources for face-to-face surveys. Knowledge gained from this research will have great impact on interviewer training and survey administration, well beyond the IAB surveys.