Eurostat – Air Transport by Aircraft and Airline (AVIA_GOEXAC / AVIA_TF series)

Two Eurostat branches that together describe who flies Europe's air cargo and on what equipment. AVIA_GOEXAC — International extra-EU freight and mail air transport between main airports and partner world region and countries — carries 10,794,784 observations of tonnage between individual EU main airports and about 200 partner countries and world regions, from 1993 through 2026-Q2. The sibling aircraft-and-airline branch adds AVIA_TF_ARC, freight and mail on board broken down by aircraft model, and AVIA_TF_LIC, which splits tonnage between airlines holding an EU licence and airlines holding a non-EU licence.

One caveat worth knowing before you build on it: the code avia_go_caa that shows up in some listings does not exist in Eurostat dissemination at all. The anchor for this record is AVIA_GOEXAC, with the two AVIA_TF tables supplying the carrier-and-equipment cut the name promises.

The route numbers are the story. Frankfurt/Main to China moved 374,920 tonnes of freight and mail in 2019, dipped to 366,209 in 2020, recovered through 436,155 in 2021 and 384,814 in 2024, then reached 465,483 tonnes in 2025 — a single lane's full pandemic arc, readable straight off the rows.

Every row is a dimension combination plus a measured value. Frequency runs annual, quarterly or monthly; the unit is either tonnes of freight and mail or flight counts; and the traffic measure separates loaded, unloaded, total loaded-and-unloaded, and on-board tonnage so capacity utilisation can be derived rather than guessed. Reporting airports come from a codelist of roughly 1,130 ICAO-coded main airports, partners from a 488-code codelist spanning single countries and aggregates such as EXT_EU27_2020. Each observation carries a value and a status flag marking estimates, breakings in series and other qualifications.

The core nine fields are tabled below. A handful of dimensions appear only in specific companion tables — the aircraft model identifier in AVIA_TF_ARC, the EU/non-EU licence indicator in AVIA_TF_LIC, and the intra-EU variant dimensions used by the domestic-leg tables — and we fold those under additional fields on request rather than pretending they sit in every row.

Geographic: EU, EFTA and candidate reporting countries at main-airport level — about 1,130 coded airports — crossed with roughly 200 partner countries and world regions, so a lane like Frankfurt–China and a bloc-level view like EU27-to-North-America come out of the same dimension set.

Temporal: 1993 through 2026-Q2, at annual, quarterly and monthly frequencies. The long history is partitioned: partner-country detail from 2008 onward sits in the current extra-EU table while 1993–2007 records live in its predecessor, and we stitch the seam so one time series comes back.

Granularity: airport x partner x traffic measure x period as the base grid, with aircraft model or airline licence status available where the companion tables apply. That makes it the natural denominator layer for air-cargo work — market-share models need exactly this lane-level tonnage before any named-carrier split can be layered on top.

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

Ask for the whole 10.8-million-row table or just the lanes you price against — a single airport-pair, one partner country, or the full partner-region matrix at monthly grain. We deliver filtered extracts and full-table loads alike, normalised to your schema, with the observation-status flags preserved so you know which values are measured and which are flagged estimates. If your models want it pushed on a schedule, delivery cadence is your call; if your analysts want it queried in place, we put it next to the rest of your tables. Sample first, contract later: request rows for the lanes you care about and check them against your own forwarder invoices before anything else happens.

Air-cargo pricing analysts use the lane-level tonnage as the volume leg of rate work: when Frankfurt–China volumes swing from 374,920 to 465,483 tonnes across six years, tender benchmarks make sense only against the underlying demand curve. Network planners at integrators and forwarders read loaded-versus-unloaded splits per airport pair to find structural imbalances worth repositioning capacity against.

Aerospace and leasing analysts go to the aircraft-model table, where freight on board by model turns fleet composition questions — how much of Europe's lifted tonnage rides on freighter types being retired — into arithmetic. Trade economists use the partner aggregates to connect customs-value flows with physical air tonnage and test whether high-value goods really travel by air at the margins. Quant teams treat the monthly series as a coincident indicator of European goods trade, ahead of slower official trade releases.

Within our air-freight catalogue of 18 datasets, this one occupies a specific corner: it is the European, official-statistics view with the deepest consistent history — 1993 onward — and the only one that reaches aircraft-model detail. Against the alternatives: BTS's T-100 series gives named-carrier granularity but only for carriers touching US points; ACI World covers airport totals without partner attribution; TAC Index prices lanes without volumes; WorldACD sells billed weight and chargeables at market level. None of those stitches airport-to-partner-country tonnage at monthly grain for three decades.

The honest limitation cuts the other way: no Eurostat table names individual airlines, because EU returns aggregate by licence status. If your question is which carrier carried what, start with T-100; if it is how much flew between which European airport and which country on what kind of aircraft, this is the record that answers it.

New reference periods arrive continuously through the year, with the latest quarters landing shortly after each month closes — 2026-Q2 figures were already in the table when we catalogued it. Three things deserve respect in the rows. First, observation flags: some cells are estimates or break in series, and treating a flagged estimate as a hard number is the classic way to corrupt a lane trend. Second, the 2008 seam between the current and predecessor extra-EU tables, which we bridge but which raw pulls would show as two different series. Third, dissemination thresholds: route-level cells can be suppressed where a single carrier dominates a lane, so thin routes sometimes report less often than headline ones.