SMVaddadi/propensity-data-eurostat-electricity

eurostat_electricity — mirror

A convenience mirror of a public dataset, pushed by propensity.data.hf_mirror so that benchmark runs do not depend on a flaky upstream origin.

This is a mirror, not a grant of rights

Copyright and licence remain with the original source named below. Mirroring moves bytes; it does not and cannot change the terms they arrived under. If you intend to redistribute this further, or to use it commercially, read the original terms — not this card, and not the license: field above, which is our best reading and may be wrong.

Provenance

• Source: Eurostat, nrg_pc_204: Electricity prices for household consumers, bi-annual (DOI 10.2908/NRG_PC_204). Source: Eurostat. Reuse for commercial or non-commercial purposes is authorised provided the source is acknowledged, read in prose from Eurostat's own copyright notice. WHAT IS PUBLISHED HERE IS A NARROWED SLICE, and the narrowing is deliberate. The published file is the commercial_only + EUR slice: 30,613 rows over 41 geographies [measured], NOT the full 83,925-row / 43-geography table the loader returns locally. Two reasons. (1) LICENCE: Eurostat bars COMMERCIAL reuse of data for countries outside the EU / EFTA / official acceding-and-candidate set and requires a commercial reuser to eliminate those rows -- here UK (1,782 obs, 1991-S1..2020-S1) and XK/Kosovo (870 obs, 2013-S1..2026-S1), 2,652 rows or 3.16%. Those rows are NOT re-hosted, because passing an elimination duty to anonymous downloaders is how it gets missed; the loader keeps them flagged for local non-commercial use, which is unrestricted. (2) CORRECTNESS: Eurostat publishes most observations THREE TIMES, once per currency, and NAC is a different unit per country while being numerically IDENTICAL to EUR for euro-area countries -- so value-level de-duplication cannot find it. Publishing the unfiltered frame would publish a known triple-count. WHY IT IS HERE: of the open-licensed price sources in this collection it has the cleanest combination of a real licence, a real time index, a real entity id and real multi-target structure -- fix geo x time and a block of (band x tax) price cells hangs off one observation with no joining. CAVEATS, all measured. (a) The honest multi-target width is NOT the 99 that 11 bands x 3 tax levels x 3 currencies suggests: the 3 currencies are the same price re-expressed, the 3 tax levels are NESTED inclusiveness levels of the same kWh, and the 11 bands are 5 + 5 mutually exclusive scheme-era bands plus 1 aggregate over them. The real simultaneous width at one (geo, time) is 5 bands x 3 tax levels = 15. Treating it as 99 feeds a multi-output model up to 85 mechanically redundant columns. (b) nrg_cons mixes a pre-2007 and a from-2007 band scheme that BOTH report 2007-S1 and 2007-S2 for 30 geographies (3,273 rows), so stacking all band codes double-books those semesters; use band_scheme=. TOT_KWH is a weighted aggregate OVER the bands, not a band. (c) geo contains two Eurostat aggregate pseudo-entities, EU27_2020 and EA, flagged by geo_is_aggregate -- and EA's membership CHANGES along the time axis (EA11-1999 through EA21-2026), so it is not a fixed entity. (d) 2,630 observations carry Eurostat status letters (1,310 provisional, 1,185 estimated, 135 definition-differs) kept verbatim in obs_status; not-available is ABSENCE, not a null. (e) 102 values are <= 0 (EUR min -0.9181) because prices include subsidies from 2021. (f) pandas cannot parse the published time codes at all -- pd.to_datetime('1985-S1') raises and errors='coerce' returns NaT for all 77 -- so period_start and time_ordinal are derived. (g) freq, siec and unit are constant columns. WHAT THIS IS NOT: not customer-level or contract-level, and NOT labelled -- there is no target and none is manufactured, so label_columns is ['value']. The entity is a COUNTRY, so an entity-disjoint split has n=41 here. DATA MODIFIED: reshaped from Eurostat's hypercube to long, with year, semester, time_ordinal, period_start, band_scheme and two flag columns derived, and restricted geographies removed. Eurostat is not responsible for these modifications and does not endorse them.
• Origin URL: https://ec.europa.eu/eurostat/databrowser/view/nrg_pc_204
• Loader: propensity.data.public_loaders.load_eurostat_electricity
• Extract: full pull from origin
• Reported source: eurostat_electricity
• Rows pushed: 30,613
• Columns: 23

Licence and redistribution

• Licence as found: other (eurostat-reuse-noncommercial-carveout)
• Redistribution verdict: open
• Basis: Eurostat, reuse authorised with attribution. Read 2026-11-10 in prose from the rights holder at https://ec.europa.eu/eurostat/web/main/help/copyright-notice (HTTP 200, 177,504 bytes). Grant, verbatim: "Reuse of statistical data, metadata, publications, and other dissemination tools published on this website for commercial or non-commercial purposes is authorised provided the source is acknowledged. The reuse policy of the European Commission is implemented by the Decision of 12 December 2011." THE CARVE-OUT, QUOTED IN FULL because quoting only the permissive half is what this table exists to refuse -- same page, under "Exceptions": "The following Eurostat data and documents may not be reused for commercial purposes, but non-commercial reuse is possible without restriction: [...] Data for countries other than: Member States of the European Union (EU); Member States of the European Free Trade Association (EFTA); official EU acceding and candidate countries. Examples are data for the United States of America, Japan or China. In such cases, the user will need to eliminate these data from the tables before reusing them commercially." SCOPE MEASURED, not assumed: in nrg_pc_204 that reaches exactly 2 of 43 geographies -- UK (1,782 obs, 1991-S1..2020-S1) and XK/Kosovo (870 obs, 2013-S1..2026-S1) -- i.e. 2,652 of 83,925 rows, 3.16%. Neither of the two aggregates present (EU27_2020, EA) contains UK data, EU27_2020 being the post-withdrawal 27 by definition and the euro area never having included the UK, so no 'an aggregate computed by Eurostat is Eurostat's data' argument is needed here (it IS needed in the sibling eurostat_arrears, whose EU28 and EU27_2007 do contain UK data). TWO FURTHER OBLIGATIONS TRAVEL, both quoted from the same page: "When reuse involves translations of publications or modifications to the data or text, this must be stated clearly to the end user of the information. A disclaimer regarding the non-responsibility of Eurostat shall be included." This loader reshapes the published hypercube to long and derives six columns, so that IS a modification and both the citation and the card state it and disclaim Eurostat. Plus the dataset citation form: Source: Eurostat, dataset nrg_pc_204, DOI 10.2908/NRG_PC_204. HOW THE MIRROR HONOURS THE CARVE-OUT: _MIRROR_LOADER_DEFAULTS publishes the commercial_only=True slice, so the 2,652 restricted rows are NOT re-hosted. The loader keeps them flagged for local use, where non-commercial reuse is unrestricted. Relying on a card to pass the elimination duty to anonymous Hub users is how that obligation gets missed.

The permission is per-row, and the frame says which rows

Eurostat authorises reuse "for commercial or non-commercial purposes [...] provided the source is acknowledged", but excepts "Data for countries other than" EU Member States, EFTA members and official acceding/candidate countries, which "may not be reused for commercial purposes" -- and states that "the user will need to eliminate these data from the tables before reusing them commercially".

So the permission is a property of the (dataset, geo) pair. Nothing has been dropped here, because non-commercial research may use all of it without restriction and a commercial reuser needs to know which rows to remove. The commercial_reuse_restricted column carries that: filter it to 0 before any commercial reuse. It fails closed, so an unrecognised geo is flagged rather than waved through.

Citation

Eurostat, nrg_pc_204: Electricity prices for household consumers - bi-annual data (DOI 10.2908/NRG_PC_204), accessed via the dissemination API. Source: Eurostat. Reuse for commercial or non-commercial purposes is authorised provided the source is acknowledged, EXCEPT that data for countries other than EU Member States, EFTA Member States and official EU acceding and candidate countries may not be reused for commercial purposes and must be eliminated from the tables before commercial reuse -- in this table that is UK and XK, marked commercial_reuse_restricted. The data has been modified here (hypercube reshaped to long format; year, semester, time_ordinal, period_start, band_scheme and the two flag columns derived); Eurostat is not responsible for those modifications.

Loading

from propensity.data.hf_mirror import load_mirrored

result = load_mirrored("eurostat_electricity", org="SMVaddadi")
result.frame.head()

load_mirrored falls back to the original loader (origin, then its own offline replica) if this mirror is unreachable, so it is safe to depend on.