Living Inflation Data Home › Sources › statfin_finland HBS AI Institute

Statistics Finland

As built

DatasetConsumer price indices
Dataset idstatfin.cpi
Series12,255
Observations1,838,186
Countries1
Period range1860-01-01 to 2026-09-01
Vintages2
Last fetched2026-10-14 10:46
Poll interval1 days
Archivestatfin/consumer-price-index/*.json.gz
LicenceCC-BY-4.0
Redistributablewith-attribution
Terms assessed2026-10-08
AttributionSource: Statistics Finland, consumer price index. Living Inflation Data transforms Statistics Finland's open PxWeb tables under CC BY 4.0.

https://stat.fi/en/about-us/get-to-know-statistics-finland/legislation/terms-of-use

What building it taught us

The connector's own docstring, verbatim. It lives beside the code so it cannot drift from it.

Statistics Finland consumer-price connector.

Statistics Finland exposes an English PxWeb catalogue with 17 current tables and a separate 39-table archive. This connector retains the price families needed for the public collection contract:

* the current 2025-base CPI, HICP and HICP-at-constant-taxes catalogues; * the source's separate 2015-, 2010- and 2005-base detailed catalogues; * older CPI main-heading families and the 1914, 1938 and 1951 cost-of-living families; and * the published CPI and HICP basket weights from 2016 onward.

Annual duplicates, preliminary HICP, OECD analytical aggregates, impact contributions and average-price tables are catalogued but deliberately not ingested. No index base or classification is spliced. The current source table explicitly backcasts its 2025-base classification to 1995; that remains one source-published family rather than a connector reconstruction.

Blank source cells remain blank. They produce two early headline gaps in the 1914 cost-of-living family, a 1962-65 gap in the retired no-tax 1938 family, three isolated historic-main-heading months, and finite gaps in seven current CPI and two current HICP categories. The archived 2010-base table also publishes six April 2021 rates that differ from their own index ratios by more than one percentage point (worst 3.61 points). The connector preserves both published quantities and does not replace either with a computed value.