Representativeness in business surveys and linked employer employee data

Abstract

A central goal of the dtec.bw project SOEP-LEE2 is to provide high-quality data for research in organisational, business, and economic studies. To this end, the project conducted two multi-year surveys among German establishments. The first survey, SOEP-LEE2-Core, is linkable to employee data from the Socio-Economic Panel (SOEP) and is characterized by a complex sampling procedure. The second survey, SOEP-LEE2-Compare, lacks this linkage possibility and has a simpler sampling design. In this article, we evaluate the sampling designs of the two surveys in term of their statistical representativeness. Our results indicate that the more complex sampling design of SOEP-LEE2-Core is susceptible to selective participation in some of the analysed dimensions, resulting in an overly large proportion of employees from the public sector. In contrast, the simpler design of SOEP-LEE2-Compare is less prone to this type of selection and yields more representative data.