dataseries: Switzerland's Data Series in One Place

Download and import open Swiss economic time series from dataseries.org, a comprehensive and up-to-date collection of public data from Switzerland. The package talks to the public dataseries.org API and imports series as pandas DataFrames.

An R package with the same interface is available on CRAN.

Installation

pip install dataseries

Requires Python ≥ 3.9 and pandas.

Data model

Data on dataseries.org is organized into datasets. A dataset is a family of related series and, in most cases, a multi-dimensional cube — a single time series is one cell of that cube, addressed by the dataset plus one code per dimension. For example the GDP dataset (ch_seco_gdp) splits along three dimensions: type (nominal/real/…), structure (GDP, value added, …) and seas_adj (seasonally adjusted or not).

• ds_catalog() lists every dataset.
• ds_search(pattern) is a flat, searchable list of the individual series.
• ds_meta(id) describes a dataset's dimensions and the codes within them.
• ds(id, ...) downloads series.

Usage

import dataseries

# Browse what's available
dataseries.ds_catalog()

# Find a specific series across all datasets
dataseries.ds_search("unemployment")

# A dataset's dimensions and codes
dataseries.ds_meta("ch_seco_gdp")

# Whole dataset (long DataFrame)
dataseries.ds("ch_fso_cpi")

# One series: pass dimension codes as keyword arguments
dataseries.ds("ch_fso_cpi", item="100_100")

# Several series, restricted to a date range
dataseries.ds("ch_fso_cpi", item=["100_100", "100_1"], start="2020-01-01")

# One cell of a multi-dimensional cube, as a wide DataFrame indexed by date
dataseries.ds("ch_seco_gdp", type="real", structure="gdp", seas_adj="csa",
              wide=True)

The long format (the default) has the dimension column(s), then date (datetime) and value (float) — ready for groupby, seaborn or plotly. wide=True pivots to one column per series with a DatetimeIndex, the shape you want for .plot() or statsmodels. All series are regular (annual, quarterly or monthly); convert with e.g. .to_period("Q") if you prefer a PeriodIndex.

Dimension arguments are optional: omit them and you get the whole dataset. Filtering happens on the server, so selecting one series does not download the whole cube. Downloads are cached in memory for the session; dataseries.cache_clear() forces a fresh download.

Labels in German, French or Italian

The catalog and search index are translated. Pass lang to get titles and labels in any Swiss national language (falls back to English where a translation is missing):

dataseries.ds_catalog(lang="de")
dataseries.ds_search("arbeitslosigkeit", lang="de")

From search hit to data

ds_search() returns exactly the columns you feed back to ds():

hits = dataseries.ds_search("consumer price")
row = hits.iloc[0]
df = dataseries.ds(row["dataset"], {row["dim"]: row["code"]})

Beyond Python

Every series is also available as a plain CSV from any tool that can read a URL:

https://api.dataseries.org/series.csv?dataset=ch_fso_cpi&dims=item=100_100

Self-hosting or testing against a mirror? Point the package elsewhere with the DATASERIES_API environment variable.

License

MIT. The data itself is published by the Swiss data providers under their respective terms; see the license column in ds_catalog() and the source links in ds_meta().