Interviewer effects and reliability of retrospective event histories within the German Generations and Gender Survey

Lenore Sauer, Federal Institute for Population Research, Germany
Kerstin Ruckdeschel, Bundesinstitut für Bevoelkerungsforschung
Robert Naderi, Federal Institute for Population Research, Germany

The life-course perspective is crucial for the explanation and understanding of demographic behaviour – especially of family formation. Therefore one of the main objectives of the Gender and Generations Programme (GGP) was to provide researchers with the information on individuals’ life course. There are two main approaches to gather this kind of information: either a prospective approach including a panel design or retrospective questions. Both possibilities have been applied in the GGS, but obviously, in the first wave only the retrospective part could be used for event analyses. However, there are known problems concerning the quality of retrospective data, especially a full recollection of past events. In the German national sample such problems evolved in the retrospective part of the survey dealing with fertility histories and marital status. The results differ considerably from vital statistics and the state of the art. Childlessness is generally and especially in older cohorts overreported and the results concerning ever married women are implausible - their share in older cohorts is too low. In this paper we investigate different reasons for this problem. First of all we identify the main distortions and the most affected groups. We then try to identify possible sources of distortions: the sampling procedure, the instrument and the interview situation. We are using several sources of comparison for this undertaking: GGS data of other countries and a second supplementary GGS sample of Turkish nationals living in Germany. Special attention is given to an extensive interviewer control taking into account several kinds of “learning effects” to shorten parts of the interview which could result in the contamination of the data. Finally we discuss possible solutions in order to provide the users of the German GGS with recommendations how to deal with these problems when analysing related questions.