About dataseries
dataseries brings Switzerland's official economic statistics together in one place. Search for a series, chart it, download it, or pull it into R or Python. The data stays under each provider's own licence (see Licenses).
The catalog draws on four providers: the Swiss National Bank (SNB), the KOF Swiss Economic Institute at ETH Zurich, the Federal Statistical Office (FSO), and the State Secretariat for Economic Affairs (SECO). Between them they cover national accounts, prices, the labour market, money and banking, interest and exchange rates, the balance of payments, business surveys, and tourism.
Where it comes from
dataseries began as an independent project and is now a fresh rewrite at cynkra. The old Shiny stack and R-only format are gone, but the idea is unchanged: take the well-maintained Swiss sources and make them easy to find, compare, and use.
The simple format behind it (a CSV for the data, a JSON for the metadata) grew out of swissdata, a joint effort with the KOF Swiss Economic Institute at ETH Zurich and cantonal statistical offices to make public data machine-readable without specialised infrastructure.
What it is for
The data already exists and is well kept by SNB, KOF, SECO and FSO. The friction is in getting at it: four APIs, four formats, four ways of naming the same thing. dataseries removes that and serves everything through one interface.
It does not invent a new data standard or try to host the world's data. For international series it points to FRED and other open sources.
How it is built
The whole pipeline that fetches, reshapes and publishes the data is public. Two repositories do the work:
- dataseries-data (public): the ETL code, the data files (CSV, JSON and Parquet), the catalog, and the format contract. Everything you can download is produced here, in the open.
- dataseries.org: the JSON API and this website. The API is a thin, read-only query layer built with plumber2 and DuckDB; the website is written in TypeScript and bundled with Vite.
Small client libraries sit on top so you never touch the raw files by hand. The R package revives the original dataseries name on CRAN, and a Python package with the same interface is on PyPI. Both read the same data, so a number matches whether you pull it in R, Python, or as a CSV.
Want a series we don't have yet, or spotted something off? Open an issue on GitHub.
Thanks
Many people have contributed to dataseries and its predecessors over the years. Special thanks to Matthias Bannert for all the thinking and collaboration on swissdata; and to Severin Thöni, Thomas Knecht, Jannes Muenchow and Karolis Koncevičius for their contributions to earlier versions of dataseries.