Eurostat Waste statistics (generation & treatment)

Europe's official answer to a deceptively simple question: what do we throw away, and where does it go? The collection is Eurostat's waste theme - roughly 30-40 harmonised tables that compile the EU's cross-country statistics on waste generation and management for the member states, EFTA countries and candidates. Where most national environmental agencies publish their own formats on their own schedules, this is the version where Lisbon, Helsinki and Warsaw sit on identical definitions.

Six streams carry most of the analytical weight: [municipal waste](#) (env_wasmun), waste generated by economic activity and hazardousness under NACE Rev. 2 (env_wasgen), [packaging waste](#) (env_waspac), waste electrical and electronic equipment (env_waselee) and waste batteries (env_wasbat), rounded out by indicators on lightweight plastic carrier bags and waste shipments between countries. Each stream splits by management operation - generation, then the fork between disposal (landfill, incineration without energy recovery) and recovery (recycling, composting, incineration with energy recovery, backfilling).

Datadory delivers this as a production feed: the full harmonised panel shaped for whatever your models, dashboards or disclosure templates expect.

One observation per row: a collection, a geography, a year and the reading. Straight from the published collections:

dataset=env_wasmun geo=EU27_2020 unit=KG_HAB wst_oper=GEN
  time=2022 value=511     (EU-27 municipal waste generated, kg per capita)
  time=2023 value=517     (next reference year - six kilos more per person)

dataset=env_waspac geo=EU27_2020 unit=KG_HAB waste=W150102 wst_oper=GEN
  time=2023 value=35.28   (plastic packaging generated, kg per capita)

Three facts hide in five numbers. Municipal generation ticked up from 511 to 517 kg per capita between 2022 and 2023 - the EU's per-person waste footprint grew even as recycling policy tightened. The packaging table drills below the aggregate to individual material codes, W150102 being plastic packaging at 35.28 kg per capita. And every value arrives with a companion flag distinguishing estimated from confirmed figures, which is the difference between a defensible chart and a provisional one.

Eight fields define every observation, and each is a documented dimension code rather than free text - which is why joins never drift between years or geographies. The dictionary below is the complete core schema; nothing here is inferred.

Two fields deserve a second look. wst_oper is the analytical spine: GEN, RCY, COM, DSP_I and related codes turn each stream into a full mass balance rather than a single headline number. And the observation flag travels beside every value - e for estimated, p for provisional, b for a break in series, u for low reliability - so data quality is a column you can filter on, not a footnote you have to hunt for.

Beyond the schema above, derived columns ship as additional fields on request: computed recycling-rate ratios, country groupings, two-level stream rollups and crosswalks between waste codes and national reporting categories.

Geography - EU-27 member states plus EFTA and EU candidate countries, with EU and euro-area aggregates sitting beside the national rows. Aggregates and countries ship in one structure, so a Europe-wide chart and a per-country breakdown come from the same query shape.

Temporal - municipal waste series reach back to 1995, giving three decades of per-capita history; most other streams start in the mid-1990s or 2000s and run to the latest completed reference year. Publication trails reality by design: municipal waste for 2024 was released in March 2026. Treat the feed as confirmed history rather than a nowcast.

Granularity - annual national-level observations disaggregated four ways at once: by waste stream and its statistical classification, by management operation, by hazardousness class, and - inside env_wasgen - by NACE Rev. 2 economic activity. The largest single table returns about 22 MB unfiltered, mid-sized tables 1-2 MB. There is no sub-national or company-level detail; for facility-level US handling records, EPA ECHO covers that job on the other side of the Atlantic.

API, files, or your warehouse. Daily, weekly, or hourly.

Pick the cadence and the landing zone; the same harmonised rows arrive either way, filtered to whichever streams, operations and geographies your work names. The sample comes first - rows and field dictionary confirmed against your use case before any commitment.

• Circular-economy and packaging disclosure - anchor ESG reporting on harmonised per-capita generation and recycling shares instead of reconciling twenty-plus national publications with different boundaries and vintages.
• Policy evaluation - test whether landfill and incineration shares actually bend after directive deadlines, using operation codes that separate disposal from recovery rather than self-reported national summaries.
• Market sizing for operators and investors - convert thousand-tonne tables into addressable treatment volumes per country, then overlay per-capita trends to see which markets are growing in volume versus merely in population.
• Forecasting - train on an annually refreshed, dimension-coded panel whose flag column tells the model which points are estimated; three decades of municipal history spans several policy regimes.
• Regulatory intelligence for producers - track packaging and WEEE streams that drive producer-responsibility obligations, watching the material codes that regulation targets rather than broad aggregates.
• Reporting and research - cite a DOI-pinned series for Europe's waste story instead of stitching press releases.

ESG and sustainability analysts get disclosure-grade harmonisation - one definition of recycling across the entire EU-27, which no aggregation of national reports achieves. Policy and regulatory researchers get the operation-level split needed to evaluate directives rather than describe them. Waste-sector investors and operators get tonnage and per-capita views of every market they might enter. Data scientists and ML engineers get a schema-stable panel long enough to forecast on, with data-quality flags built into the rows. Journalists and academics get one citable, DOI-backed source instead of a bibliography of national statistical offices.

Three things worth knowing upfront. First, publication lags completion: the newest municipal waste figures described calendar 2024 and arrived in March 2026, so build pipelines around confirmed history and pair the feed with a leading indicator if you need nowcasting. Second, the collection counts tonnes and kilograms, not euros - market sizing in currency needs a price layer joined separately, and we can discuss that join when you scope your sample. Third, series breaks are real: classification changes over three decades leave b flags in the history, and any long-run trend should be checked for them before it becomes a slide.