Ende der 1990er Jahre begann eine Diskussion zu den Möglichkeiten einer quartalsweisen europäischen Statistik über offene Stellen. Während es sehr umfängliche Informationen über die Arbeitssuchenden in den amtlichen Statistiken gibt, stehen über den Arbeitskräftebedarf nur sehr begrenzt Informationen zur Verfügung. So sind in den meisten Ländern Betriebsbefragungen die einzige Möglichkeit, die Gesamtzahl der offenen Stellen zu ermitteln. Um auf europäischer Ebene die Vergleichbarkeit der Länderstatistiken gewährleisten zu können, müssen vergleichbare Definitionen und Methoden verwendet werden. Im Jahr 2005 formulierte die Europäische Kommission zusätzliche und sehr spezifische Anforderungen zur Aufsplittung der Daten nach Regionen (NUTS), Wirtschaftszweigen (NACE) und Berufen (ISCO). Mit unserer Arbeit versuchen wir die Frage zu beantworten, ob es möglich ist, Daten auf tief disaggregierter sektoraler, regionaler und beruflicher Ebene zu schätzen. Es werden Mechanismen, die zu fehlenden Werten führen, ebenso diskutiert, wie die Stärken und Schwächen der gewöhnlich verwendeten Imputationsmethoden.

"At the end of the 1990s a discussion on the need and possibilities to set up quarterly European statistics on unmet labour demand as a counterpart to the unemployment statistics on the supply side of the labour market has started. In most countries business surveys are the only way to find out the total number of vacancies. They need to use comparable methods and definitions to allow country comparisons. In 2005 the European Commission formulated additional and very specific requirements concerning breakdowns by regions (NUTS), sectors (NACE) and occupations (ISCO). With our work we tried to answer the question, if and how it could be possible to estimate data on a very deep disaggregated sectoral, regional and occupational level respectively. We have analyzed the available database with several methods, like Small Area Estimation, Multiple Imputation and regressions. Part of our research was successful; some questions require more research to find satisfying answers and we continue working on these. Data on occupations by NUTS 2, e.g. can not be produced without a considerable increase in the sample size and great effort in developing usable estimation techniques, since NUTS 2 regions in Germany are relatively small. A cost-benefit-analysis does not justify the large amount of financial resources required. Therefore NUTS 1 is the appropriate regional level for vacancy data from our perspective. A publication of occupational data by NUTS 2 is insufficient anyhow for good analyses of mismatch, since a classification by ISCO major or sub major groups is too rough for this. Furthermore such analyses would need to take into account regional specifics on both sides of the labour market. Missing values are a common problem in analyses of business data. Our report has discussed concepts regarding mechanisms that create missing data, as well as the strengths and weaknesses of commonly used approaches. Using only the long questionnaire from 2005, our simulation studies show that multiple imputation for the German job vacancy survey can lead to unbiased results for the missing data and multiple imputation has been used successfully for business data in several other applications. The methodological work on methods of multiple imputation and of split survey designs will be continued to reduce the answering burden for respondents and to fill in missing variables." (Author's abstract, IAB-Doku)