Comparing administrative and survey data: Is information on education from administrative records of the German Institute for Employment Research consistent with survey self-reports?

摘要

In research on stratification and inequality, administrative data are popular for their wide coverage and assumed high quality. Yet, the quality of the data depends crucially on the aim of data collection. In this paper, we investigate the quality of information on education in administrative data from social security records provided by the German Federal Institute for Employment Research where education was not the primary purpose of data collection. We use linked German employee data with self-reported education as a benchmark to investigate whether the level of education is consistent or provided at all in the administrative data. The results show striking differences between administrative and survey data. Not only is information on education often missing from the administrative data; the information contained often deviates from the information employees reported in the survey. Information on school-leaving certificates is more often missing from the administrative data than information on vocational and university degrees. Furthermore, the information on vocational and university degrees is frequently inconsistent. Our results, moreover, reveal that missingness and inconsistency of information differ by type of degree obtained. Employer characteristics show a systematic correlation with missingness of information on both schooling and vocational degrees but appear less relevant in explaining inconsistencies. Additional analyses of estimated returns to education indicate that misreporting of vocational degrees in particular leads to an underestimation of actual returns to education. These results suggest that further research on the quality of measures of education in administrative data collected for different purposes is needed.