Getting started with dataseries
What is dataseries.org?
Switzerland publishes a lot of official statistics, but they are
spread across many providers: the Federal Statistical Office, SECO, the
National Bank, the Federal Finance Administration and others. Each has
its own portal, its own file formats and its own update schedule.
dataseries.org collects them in
one place. It tracks the sources, harmonises them into a single
structure and keeps them current. This package is a client for its
public API, so you can pull any of those series straight into R.
How the data is organised
Data comes in datasets. A dataset is a family of
related series and is usually a multi-dimensional cube: one
time series is a single cell of that cube, addressed by the dataset plus
one code per dimension.
Four functions cover the whole package:
ds_catalog() lists the datasets.ds_search() lists the individual series, so you can
grep for one.ds_meta() describes a dataset’s dimensions and their
codes.ds() downloads.
Finding something
Start with the catalog. One row per dataset:
cat <- ds_catalog()
nrow(cat)
#> [1] 70
head(cat[, c("id", "title", "frequency", "n_series")])
#> id title frequency
#> 1 ch_adecco_sjmi Adecco Group Swiss Job Market Index quarterly
#> 2 ch_ffa_finances Government finances annual
#> 3 ch_fso_besta Jobs by economic division quarterly
#> 4 ch_fso_besta_outlook Employment outlook quarterly
#> 5 ch_fso_construction_prices Construction prices semi-annual
#> 6 ch_fso_cpi Consumer prices (detailed basket) monthly
#> n_series
#> 1 4
#> 2 387
#> 3 60
#> 4 20
#> 5 3
#> 6 595
If you know roughly what you want, ds_search() is finer
grained. It returns one row per series across all datasets:
hits <- ds_search("unemployment rate")
head(hits[, c("dataset", "dim", "code", "label")])
#> dataset dim code label
#> 1 ch_fso_unemp_rate origin tot Total
#> 2 ch_fso_unemp_rate origin ch Swiss nationals
#> 3 ch_fso_unemp_rate origin ex Foreign nationals
The dataset, dim and code
columns are exactly what ds() expects, so a search result
can be fed straight back in.
Downloading
The simplest call takes a dataset id and returns every series in it,
in long format:
cpi <- ds("ch_fso_cpi")
head(cpi)
#> item date value
#> 1 100_1 1982-12-01 70.0205
#> 2 100_10 1982-12-01 34.7833
#> 3 100_100 1982-12-01 59.1566
#> 4 100_1001 1982-12-01 69.2540
#> 5 100_1002 1982-12-01 65.3241
#> 6 100_1010 1982-12-01 82.6664
To pick one series, pass the dimension codes as named arguments.
Filtering happens on the server, so this does not download the whole
cube first:
total <- ds("ch_fso_cpi", item = "100_100", from = "2015-01-01")
head(total)
#> item date value
#> 1 100_100 2015-01-01 93.5394
#> 2 100_100 2015-02-01 93.2989
#> 3 100_100 2015-03-01 93.6033
#> 4 100_100 2015-04-01 93.4335
#> 5 100_100 2015-05-01 93.6510
#> 6 100_100 2015-06-01 93.7198
Working with cubes
Which dimensions does a dataset have? ds_meta() tells
you:
m <- ds_meta("ch_seco_gdp")
unlist(m$dim_order)
#> [1] "type" "structure" "seas_adj"
names(m$dimensions$type$levels)
#> [1] "nom" "real" "gc_q" "gc_y"
So the GDP dataset splits three ways, and a single cell needs one
code from each:
gdp <- ds("ch_seco_gdp", type = "real", structure = "gdp", seas_adj = "csa")
tail(gdp)
#> type structure seas_adj date value
#> 180 real gdp csa 2024-10-01 200725.0
#> 181 real gdp csa 2025-01-01 202350.7
#> 182 real gdp csa 2025-04-01 202640.5
#> 183 real gdp csa 2025-07-01 201781.0
#> 184 real gdp csa 2025-10-01 202218.0
#> 185 real gdp csa 2026-01-01 203544.3
Time series objects
Every series on dataseries.org is regular (annual, quarterly or
monthly), so it maps cleanly onto R’s ts class. Pass
class = "ts":
gdp_ts <- ds("ch_seco_gdp", type = "real", structure = "gdp", seas_adj = "csa",
class = "ts")
plot(gdp_ts, main = "Swiss real GDP, seasonally adjusted",
ylab = "CHF million", col = "steelblue")
Select several cells and you get an mts with one column
per series:
two <- ds("ch_fso_cpi", item = c("100_100", "100_1"), from = "2020-01-01",
class = "ts")
head(two)
#> 100_1 100_100
#> Jan 2020 96.2727 94.0756
#> Feb 2020 96.3427 94.1922
#> Mar 2020 97.0131 94.2645
#> Apr 2020 97.6975 93.9216
#> May 2020 98.3081 93.9610
#> Jun 2020 99.2729 93.9806
For xts, convert in one line:
xts::as.xts(ds("ch_fso_cpi", item = "100_100", class = "ts")).
Labels in other languages
The providers publish their labels in German, French and Italian as
well, and dataseries.org keeps them. Pass lang to
ds_catalog() or ds_search():
head(ds_catalog(lang = "de")[, c("id", "title")])
#> id title
#> 1 ch_adecco_sjmi Adecco Group Swiss Job Market Index
#> 2 ch_ffa_finances Öffentliche Finanzen
#> 3 ch_fso_besta Beschäftigte nach Wirtschaftsabteilung
#> 4 ch_fso_besta_outlook Beschäftigungsaussichten
#> 5 ch_fso_construction_prices Baupreise
#> 6 ch_fso_cpi Konsumentenpreise (detaillierter Warenkorb)
Searching matches the labels in the chosen language, so German terms
work directly:
ds_search("arbeitslosen", lang = "de")[, c("dataset", "code", "label")]
#> dataset code label
#> 1 ch_seco_concon ks_i32_unemp_exp_q 3.2 Entwicklung der Arbeitslosenzahl
Where a translation is missing, English is used.
Caching
Everything downloaded is cached in memory for the session, so
repeating a call costs nothing. cache_ls() shows what is
held, keyed by request URL, and cache_rm() empties it to
force a fresh download:
length(cache_ls())
#> [1] 7
cache_rm()
length(cache_ls())
#> [1] 0
Beyond R
The same data is available from Python through the dataseries package
(pip install dataseries), and as plain CSV from any tool
that can read a URL:
https://api.dataseries.org/series.csv?dataset=ch_fso_cpi&dims=item=100_100
