Accessing georeferenced data
Accessing georeferenced SOEP data
The GeoLAB team and the Leibniz ScienceCampus are working towards secure remote access to georeferenced SOEP data from anywhere in the world. Until that access is in place, regional SOEP data is available through DIW Berlin: the SOEP’s page on regional data sets out the routes, the conditions and the application forms.
Finding other georeferenced data
Two search tools help you find out which variable or indicator exists, and which institution publishes it. Both search metadata: you describe a concept in plain language, get ranked descriptions back, and every hit links to the institution that publishes the data. Neither tool stores or serves measurement values or microdata.
Georeferenced regional data
GeoDB
GeoDB is short for GeoDataBase. It holds regional indicators, statistical tables and datasets for Germany: INKAR, the table catalogues of the Regionaldatenbank and GENESIS-Online, Zensus 2022, the Regionalatlas, the Deutschlandatlas, indicators of the Bundesagentur für Arbeit, the Breitbandatlas including its grid-cell rasters, hospital registers, the Unfallatlas, public-transport datasets and more. GESIS adds German and international study metadata and documented geographic variables. These can describe individual surveys rather than regional aggregates; the finder does not provide restricted survey data or promise a ready-made map join.
Ask for a concept and filter by source, spatial level, theme and year. Every result says what its link opens: the indicator itself, the table it sits in, the statistic, the dataset or the portal’s entry page.
Survey variables
SOEP Variable Finder
Variable metadata of the German Socio-Economic Panel (SOEP-Core). Ask in German or English and get the variables that measure a concept, with dataset, label, question wording, value labels and the survey years in which they appear.
Metadata only. Access to the SOEP microdata itself requires a data-use agreement with DIW Berlin; the finder tells you which variable to ask for.
How the search works
Both tools work the same way. A multilingual language model turns your question and every description into numerical vectors, and the descriptions closest to your question become the candidates. A second model then reads each candidate together with your question and puts them in order. Shared words and a small preference for core sources settle the final ranking.
The search works by meaning and is therefore imperfect: check a hit against the official documentation before you rely on it.
Improving the search
Search text is retained for quality evaluation only if you separately opt in under Privacy settings. Without that choice, only daily search totals are retained. Opt-in query records are not anonymous: free text can contain personal information. They do not contain IP addresses, user agents or the independent usage-analytics ID. See the privacy notice.
Sources and citation
The indexed sources and their attribution requirements are listed under Data sources and attribution. Both tools are open source and archived with a DOI, GeoDB as 10.5281/zenodo.21134145 and the SOEP Variable Finder as 10.5281/zenodo.21134306; each tool also shows its citation in its own footer.