World Bank Air Transport Passengers Carried (ICAO-sourced, 1970-2023)

One indicator, the whole map of commercial aviation demand. IS.AIR.PSGR sits in the World Development Indicators' infrastructure topic and reports passengers carried by airlines registered in each reporting economy, regardless of where those passengers began or ended their journey. The underlying compilation is the Civil Aviation Statistics of the World assembled by the International Civil Aviation Organization, supplemented by ICAO staff estimates where carriers do not report.

Breadth is the selling point. Roughly 265 member economies appear alongside aggregates - World, the European Union, income and regional groups - observed annually from 1970 through 2023, about 14,000 economy-year observations that fit in a few hundred kilobytes. The registered-carrier basis matters as much as the breadth: a country earns credit for the traffic its own airlines fly anywhere on earth, which makes the series a map of flag-carrier reach rather than of runway activity. That is the opposite cut from airport-throughput counts, and analysts who conflate the two produce quietly wrong comparisons. The wider shelf sits on our passenger airlines data hub.

Get a sample of this dataset - take the World recovery rows plus a few economies you actually sell into, and judge the fit before committing.

Rows arrive flat: one economy-year observation carrying the codes, the year and the count. Straight from the record - the five verified World aggregate rows:

# Verified World aggregate rows - indicator IS.AIR.PSGR
countryiso3code  country  date  value_passengers_carried
WLD              World    2023  4268984474.03796
WLD              World    2022  3230288279.667
WLD              World    2021  2279974770.123
WLD              World    2020  1771894087.0245
WLD              World    2019  4455690101.02828

Five lines, one industry's wildest decade. 4.46 billion passengers in 2019 becomes 1.77 billion in 2020 - roughly sixty percent of global demand erased in a single row-to-row step. The rebuild runs 2.28 billion in 2021, 3.23 billion in 2022 and 4.27 billion in 2023, which is about ninety-six percent of the old peak: recovered, but visibly not round. Because every row in the series carries these same four analytic columns, stacking fifty-four years across hundreds of economies builds the full panel without re-keying anything.

Thirteen documented fields carry the entire base schema, and they split into three jobs. country.id, country.value and countryiso3code do the identification - which economy, under which code, in which ISO spelling - with the ISO3 field doubling as the join key to almost any other country-level table. date, value and unit do the measuring: the reference year, the passenger count itself, and a unit field deliberately empty because the count is passengers. The rest is bookkeeping and attribution - the fixed indicator id and label, the observation-status flag that marks estimated or absent values, the display-decimal hint, and the definition and organization strings that carry the ICAO compilation credit inside every row.

Every definition in the table below was verified against delivered rows rather than inferred from documentation, so it describes what arrives instead of what a brochure promises. Refresh bookkeeping and derived columns such as growth rates and share-of-world fold under additional fields on request, documented against delivered extracts, so the base schema stays honest about what ships first.

Geography - every World Bank member economy, roughly 265 of them, plus the aggregates: World, European Union, income groups and regional groups. Each economy's figure counts passengers on its own registered carriers, so the aggregate rows are sums over registries, not over territories.

Temporal - annual observations from 1970 through 2023. Fifty-four observation years per economy where reporting holds, which reaches back further than any other historical series in the passenger-airlines slice: BTS T-100 traffic begins in 1990 and US on-time records in October 1987.

Granularity - one row per economy per year, national total. No airport, route, carrier or month-level resolution exists here by design - pair it with a flight-level or airport-level feed when the question needs that cut.

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

You pick the channel and the cadence; typing, cleaning and schema stability are our problem. Rows ship with all thirteen base fields populated and the observation-status flags preserved, so estimated values announce themselves instead of masquerading as measured ones. Bulk pulls land as files, continuous consumption runs through the API, and warehouse-native loads write straight into your own storage. A sample goes out first, cut to the economies, aggregates and year range you actually need - get a sample of this dataset and test the schema you will ship against.

• Cross-border market sizing - one definition across roughly 265 economies ranks national air-travel markets without reconciling national vocabularies, the comparison no single-country source supports.
• Pandemic-recovery tracking - the 2019-2023 rows quantify a fall to forty percent of peak and a climb back to ninety-six, in units nobody argues with.
• Long-run demand modelling - fifty-four annual observations per economy feed econometric and ML demand models that need pre-1987 history no operational archive provides.
• Flag-of-registry capacity studies - counts follow carrier registration, so alliance and fleet strategy surfaces in national series rather than hiding in airport totals.
• Forecast benchmarking - twenty-year traffic assumptions from airline and manufacturer planning decks get tested against one consistent baseline.

Market researchers and consultants (relevance 3 of 3) treat it as the canonical cross-country sizing benchmark; see market researchers use cases. Developers and data-product builders (3 of 3) embed one stable, fully documented series behind dashboards instead of maintaining dozens of country integrations; see developers builders use cases. Data scientists and ML engineers (2 of 3) ingest a ready-made economy-year panel spanning 1970 onward; see data scientists use cases. Competitive intelligence teams (1 of 3) read expanding national markets as context for carrier moves; see competitive intel product teams use cases. Journalists and academics cite it because ICAO attribution survives peer review and print style guides alike.

Against BTS TranStats T-100: the American archive digs far deeper per route - census-level monthly traffic for certificated US carriers from 1990 - but stops at US registrants and the modern era. This series goes shallower per row and wider per map: annual, global, back to 1970. Against the Airline Passenger Satisfaction survey (129,880 static responses): that set explains why individuals rate flights; this one measures how many people flew at all - we break the pairing down in vs Airline Passenger Satisfaction. OpenFlights supplies the network snapshot that turns these national totals into reachable maps. And the companion air freight series spans the identical 1970-2023 window on the same economy-year grain, so a passengers-plus-cargo panel is one extra request - see World Bank Air Transport Freight Million Ton-km Data.

Four conventions worth knowing upfront. First, this measures registries, not runways - a hub country is credited only for its own carriers, so airport-throughput questions need an airport-based feed rather than this one. Second, aggregates ride beside countries: World, EU and income-group rows use the identical schema, so filter by membership before computing any total. Third, reporting is uneven - some economies return nulls in recent years, so check completeness for your exact window before fitting trends. Fourth, values currently run through 2023, and whether newer vintages exist beyond what this catalog captured is an open question we resolve with your sample rather than guess at. For positioning against the rest of the slice, the best passenger-airlines datasets ranking puts this series second overall on comparability grounds.