Work package leader:

Michael Bergmann


Email Michael

A social survey needs to represent the population. Drawing a proper probability sample is therefore a central task.

This is often difficult due to: restricted access to sampling frames; the varying nature and quality of sampling frames; the trade-off between cost-effective deviations from simple random sampling and the size of the consequent design effects; and a climate of lower response rates. Furthermore, panel surveys require rather complex sample refreshment due to attrition. This WP brings the combined expertise of SHARE and ESS together with the EVS and GGP to develop ex ante and ex post solutions that benefit not only these infrastructures but also other cross-national social surveys in Europe. The WP has the following objectives:

  • Improve European sampling practices and examine how to make samples more inclusive and accurate
  • Find more efficient methods for accounting and correcting for non-response
  • Find economies of scale through collaborating across surveys

WP2 tasks

The WP is divided into five tasks:

Mapping and improving European sampling practice

This task reviews sampling practices across European surveys and aims to exploit the synergies and economies of scale to be gained from sharing knowledge.

Learning from administrative data

This task explores the potential for exploiting the growing amount of pre-existing administrative data to better understand and overcome survey non-response, one of the major challenges facing social surveys.

Weighting for complex survey designs

This task is to develop, document, and provide a (Stata) program code which computes calibrated weights for any set of subsamples/waves directly from the micro data.

Handling of item non-response

This task is to develop generally applicable imputation techniques to deal with item non-response.

Including the institutional population into a sample survey of the general population

This task will systematise the decisions made when including the institutional population as part of the overall sampling design and make recommendations to optimize the coverage of the institutional population.