Optimization of sampling design and inference for establishment surveys

• Applicant: Professorin Dr. Susanne Rässler (†)
• Subject Area: Empirical Social Research
• Term: from 2008 to 2012
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 43824110

Establishment surveys differ from household surveys in many important ways. One of the crucial requirements in conducting business surveys is an accurate and up-to date sampling frame. Another challenge are highly skewed variables which stem from the presence of many small but only few large business units. However, the social sciences rely on methodologically sound business surveys, as these are the only data sources for many important descriptive or econometric analyses of wage structures, corporate decision making and other business related questions.In the proposed project, we assess the impact of different weighting techniques and different variance estimation procedures in the context of business surveys. We will come up with guidelines and recommendations for "best practice" in future business surveys. The rich administrative data of the Federal Employment Agency will be used for conducting unit non-response analyses and for trying to find models for "stratum jumpers" and outliers. Additionally, we intend to find and implement improved methods for multiply imputing missing values in business surveys and to explore the possibility of "Split Questionnaire" survey designs, which promise to significantly reduce response burden.

• DFG Programme: Priority Programmes
• Subproject of: SPP 1292: Survey Methodology
• Participating Person: Professor Dr. Hans Kiesl