Comparing survey data with an official administrative population: assessing sample-selectivity in the IAB Establishment Panel

The IAB Establishment Panel is a large annual establishment survey and serves as a data source for many empirical analyzes, e.g., on labor demand, wages and industrial relations but also for influential policy evaluations. We analyze the selectivity of the IAB Establishment Panel compared with the administrative population of all establishments in Germany to test whether the IAB Establishment Panel is biased towards “good employers”. By design the survey over-samples large, thus high paying and stable establishments. After flexibly controlling for elements of the sampling design, we do not detect any meaningful difference between establishments participating in the survey and the full population regarding key economic performance indicators. If anything, we observe a slight overrepresentation of stable and slow growing establishments. For applied research, the results highlight that “typical” econometric specifications may not be sufficient to control for selectivity.