The Quality and Selectivity of Linking Federal Administrative Records to Respondents and Nonrespondents in a General Population Sample Survey of Germany

Abstract

Various forms of auxiliary information are being sought to augment survey samples and adjust for possible nonresponse bias in key survey estimates. Auxiliary data options are typically limited in most general population surveys and there are questions concerning their utility for nonresponse bias evaluation and adjustment. Federal administrative databases provide a potentially rich source of auxiliary information for nonresponse purposes, but linking them to general population samples is usually restricted to surveys which draw their samples from population registers containing unique personal identity numbers which can be directly linked to federal databases containing more detailed substantive information. In this article, we examine the quality and selectivity of augmenting a federal administrative database to a general population survey when such a unique personal identifier is not available. We employ a series of standard linkage procedures that rely instead on non-unique and error-prone identifiers collected from the sampling frame to link a federal employment database to a general population survey in Germany. The quality and selectivity of the established links are evaluated using household- and person-level interview data in accordance with German data protection laws. We report a linkage rate of 60 percent for the entire sample under a strict linkage criterion, and 80 percent under a more relaxed criterion. We find that linkage rates vary across some household- and person-level characteristics that are likely specific to the particular administrative database used in this case study. We conclude with a general discussion of the practical implications of this work for survey organizations considering performing similar linkages and highlight some opportunities for future research.