Variance Estimation using Microcensus Data under consideration of Sampling Design

The microcensus (MZ) is carried out as a stratified area sample with sample districts as a primary unit. Not all stratification variables are available in the factually anonymized microcensus (FAMZ). Moreover, the selection of the FAMZ from the MZ represents another stage of the survey. The consequence of this complex design is that no standard assumption of a simple random sample can be made for variance estimates on the basis of factually anonymized microcensus data. To date, only design-effect factors published by the Federal Statistical Office could be used for projections or statements about the population. However, these factors are not available for every unit of analysis and do not take into consideration the sampling of a 70 percent survey at household level for the factually anonymized file. The 1996 microcensus was the first time that the building category as well as identification of the primary unit (system-independent sample district number) was made available for the purposes of projection under consideration of sample design, so it was subsequently possible to examine the question in close cooperation with Prof. Dr. Ulrich Rendtel (J.W. Goethe University, Frankfurt/M.) as to how users of the factually anonymized microcensus data can adequately take into account the sample design for variance estimation. For variance estimation, programs are developed for the most common statistical packages (SAS, SPSS, Stata).

• Schimpl-Neimanns, B., 2019: Varianzschätzung von Nettoveränderungen mit dem Mikrozensus ab 2012. AStA Wirtschafts- und Sozialstatistisches Archiv 13 (1): 73-85. doi: http://dx.doi.org/10.1007/s11943-019-00240-0.
• Schimpl-Neimanns, B.; Zins, S., 2018: Estimation of the standard error for net changes with the EU Labour Force Survey - How can users independently and appropriately calculate standard errors and confidence intervals? Proceeding Paper to the European Conference on Quality in Official Statistics, Kraków, 29 June 2018, Session 29. https://www.gesis.org/fileadmin/upload/dienstleistung/daten/amtl_mikrodaten/europ_microdata/EU-LFS/paper_29_5.pdf.

• Schimpl-Neimanns, B. 2011: Schätzung des Stichprobenfehlers in Mikrozensus Scientific Use Files ab 2005, AStA Wirtschafts- und Sozialstatistisches Archiv 5 (1): 19-38. DOI: dx.doi.org/10.1007/s11943-011-0092-4 (Preprint) 

• Schimpl-Neimanns, B., 2010: Varianzschätzung für Mikrozensus Scientific Use Files ab 2005. GESIS-Technical Report Nr. 2010/03. [.pdf] (1.22 MB)

• Schimpl-Neimanns, B., 2009: Schätzung des Stichprobenfehlers im Mikrozensus mit Stata – Eine Einführung mit Beispielen zum Campus File Mikrozensus 2002. GESIS-Methodenbericht 2009/02. Mannheim: GESIS. [.pdf] (1.26 MB)

• Schimpl-Neimanns, B.; Rendtel, U., 2001: SAS-, SPSS- und STATA-Programme zur Berechnung der Varianz von Populationsschätzern im Mikrozensus ab 1996. ZUMA-Methodenbericht 2001/04. (234 kB)
  See also "Calculating sampling error in the microcensus"
• Rendtel, U; Schimpl-Neimanns, B., 2001: Die Berechung der Varianz von Populationsschätzern im Scientific Use File des Mikrozensus ab 1996. ZUMA-Nachrichten 48: 85-116. (4.60 MB)

• Rendtel, U.; Schimpl-Neimanns, B., 2000: Varianzschätzungen für den faktisch anonymisierten Mikrozensus. In: Jahrbücher für Nationalökonomie und Statistik 220/6, 2000, S. 759-776.