Reliability of Retrospective Event Histories within the German Generations and Gender Survey

Abstract

Collecting retrospective data generally is accompanied by questions concerning the quality and reliability of such data. The aim of this article is to contribute to this body of research by analysing the fertility and partnership histories within the German Generations and Gender Survey (GGS). Although landmark events such as the birth of children or marriage are commonly regarded as reliable memories and therefore are assumed to be applicable retrospectively without much decrement in data quality, we find severe distortions in the retrospective data on fertility and relationship in the German GGS by comparison to the Microcensus. In fertility histories there is a major overestimation of childless women in older cohorts as well as an underestimation of this group in middleaged cohorts. Regarding partnership history we have too many women never married in the older cohorts and too many married in the younger cohorts. We find that these distortions are mainly attributed to problems caused by survey organisation. The random route sampling procedure of the German GGS has led to problems, as has the instrument, which lacked control mechanisms while simultaneously being very complex. We also find indications of problematic interviewers, but we cannot delete the deviations in the data to our satisfaction as they are bound to be multicausal. We therefore conclude that there are combined effects of the sampling procedure, the complexity, and the length of the instrument that all contribute to interviewer effects.