The European Laeken indicators were constructed to measure poverty and social cohesion. To ensure comparability of the Laeken indicators within Europe, the European Survey on Income and Living Conditions (EU-SILC) was launched. The publication of the Laeken indicators by each EU member is necessary but not sufficient to gauge the progress towards agreed EU objectives.
As a result from the open method of coordination, the SILC sampling design varies from country to country. This urges the needs of comparable accuracy measurement methods for the SILC sampling designs.
The paper will address variance estimation methods for linear statistics, such as means and totals, as well as the non-linear Laeken indicators in the SILC context under a variety of sampling designs. The methodology comprises several linearization methods as well as the classical resampling methods. Special emphasis will be put on rare table values which in general lead to difficulties in the selection of methods. Further the problem of missing stratum information due to anonymization will be addressed. And finally, the estimation of design for Laeken indicators will be considered. To empirically explore the quality of the proposed variance estimates the results from a Monte-Carlo study will be presented which is based on a synthetic but close-to-reality dataset.
This research was part of the research project “Advanced Methodology for European Laeken Indicators” (AMELI) funded by the European Commission within the 7th Framework Program.
Keywords: Variance estimation; Non-linear statistics; Design effects; Small area estimation
Biography: Ralf Münnich is university professor and chairholder of the economic and social statistics department in faculty IV of the University of Trier. His main research interest is computational survey statistics, survey design, variance estimation, and indicator methodology. Since 2001, he has coordinated several Eureopan Research Projects supported by the european Commission. Latest publications are focused on small area statistics and specialized sampling algorithms for census methodoligies and poverty indicators.