Methods & Data
AirTraffickonomics is a study of air travel demand and of the airline business around it, built from public statistics. This page says, in outline, where the data come from, how the work is organised, what was estimated for each question and what the study does not establish.
Data sources
The United States data come from the Bureau of Transportation Statistics (traffic, freight, fares, carrier finances and fleets), the Transportation Security Administration (airport security counts), and Federal Reserve Economic Data (FRED), the Bureau of Economic Analysis and the Bureau of Labor Statistics (the economic series). The traffic data cover scheduled service only, and the financial data cover certificated carriers only.
Data vintage. The United States data run through quarter 3 of 2025: 2024 is the last full year, and 2025 is in the data only up to that quarter. The site is built on this fixed vintage: later periods enter with a dated revision, not as the agencies publish them. These periods describe the United States data only; other markets end in their own periods and state their coverage on their own charts.
Each other market takes its traffic series from its aviation or statistics authority:
| Market | Aviation source |
|---|---|
| EU (country level) | Eurostat |
| Japan | MLIT |
| UK | Eurostat (earlier), CAA (later) |
| Canada | Statistics Canada |
| Mexico | AFAC |
| Australia | BITRE |
| South Korea | Korea Airports Corporation, via KOSIS |
| Singapore | SingStat |
| Brazil | ANAC |
| China | National Bureau of Statistics, CAAC |
| India | DGCA |
The economic series come mostly from FRED, but GDP comes from Eurostat for the EU, SingStat for Singapore, the Office for National Statistics for the UK and the Organisation for Economic Co-operation and Development for China and India. The UK traffic series joins two sources: Eurostat for the earlier years, which has no UK data for the early two-thousands, and the CAA’s airport counts for the later years, where the domestic and international split begins. KOSIS is the Korean Statistical Information Service, CAAC is China’s civil aviation authority and DGCA is India’s civil aviation regulator.
How the work is organised
Raw agency files go in at one end, and every step after that is code: the files are cleaned and checked, combined into tables by quarter and by year, used to fit the models, and turned into charts. No figure is edited by hand, and the numbers in the site’s text are read from the same published data files when the site is built.
What was estimated
For each of the site’s questions: what was measured, the kind of method, and the main thing the result cannot show. The figures are on the findings page and the charts.
Do people fly more just because they get richer?
How much passenger traffic rises with national income over the long run (an elasticity), from a long-run regression of passengers on GDP, fitted the same way in each market, with a test of whether the two stay tied together over time (cointegration). Ticket prices are left out, so the figure also counts cheaper fares where they came with rising incomes, and it is a long-run figure: the response in the short run is smaller. Income charts
How much flying is there, and how fast is it growing?
Passengers, growth rates, load factors and seasonal patterns, mostly counted from the agency tables. The airport security nowcast is a ratio bridge between two counts of the same passengers, not a demand model. Each market counts passengers the way its publisher does, so the counts are not like for like across markets. Traffic charts
When does a plane leave the fleet?
How long aircraft stay in United States carrier fleets, from survival curves that count the aircraft still flying as well as those that have left. Leaving a fleet means no longer being filed by a US carrier, which includes a sale abroad as well as scrapping; a stricter definition puts every figure later. Fleet charts
Who makes money flying planes?
Cost and revenue per seat mile by carrier type, from carriers’ financial filings. Differences between carrier types describe two groups of carriers, not the effect of a business model. Airline economics charts
How much more does a longer flight cost?
How fares on United States routes vary with distance, traffic, competition and fare restrictions, from a regression that prices a route by its characteristics, on ticket samples that start in the pandemic years. The result is descriptive: fares, traffic and competition are set together on each route, and much of the difference between routes is left unexplained. Fare charts
Where does flying concentrate?
How traffic is spread across airports and routes, measured directly from the airport and route tables. The network measures count each route once whatever it carries and shift at a change in which carriers report, so no long-run trend is claimed from them. Network charts
What did COVID do, and is it over?
The fall in traffic and the recovery, by market, region and route type. Each recovery chart states its own base period, and the base periods differ. Shock charts
How much freight goes by air, and who carries it?
How much freight and mail move by air, which airlines carry them and how freight moves with the wider economy, mostly counted from the agency tables. The charts describe and do not measure what drives freight. Freight charts
The projections
How many people will be flying in 2044?
These charts are projections, not forecasts of what will happen. Outside Singapore, each market’s passengers are projected from income per person and population alone, on a curve fitted across markets in which flying rises with income and then levels off. The inputs are World Bank history, International Monetary Fund growth projections and United Nations population projections; after the IMF’s years, income growth per person is assumed to move toward one common long-run rate, an assumption with no external source. Airline expansion, new routes and low-cost entry are left out. Re-run from earlier years with actual income, on the curve fitted to all years, the projections ran below actual growth in most pre-COVID windows, and no market’s projection is called a floor. India’s and the United Kingdom’s series are too short for the re-run. Singapore, a hub, follows a time trend instead.
The scenario lines are other income and population paths and carry no confidence band. No chart on this site projects aircraft numbers. On the seat-mile what-if chart each line compounds one assumed growth rate with no external source, and it is not the site’s own projection.
Status labels
Most charts that report a model’s result or a projection carry one of these labels, with a condition where the result holds only under stated assumptions. The glossary has the full definitions.
- Confirmed: the tests support the result. For a market’s long-run link, the main test also does not fail outright on pre-COVID data alone, and where a second test dissents there, the chart says so.
- Confirmed existence: the main test confirms that a long-run link exists, but its form is not pinned down.
- Provisional: the evidence points toward the result without confirming it.
- Inconclusive: the evidence cannot settle the question. A result that is only not significant gets this label, never a finding of no effect.
- Exploratory: a restricted sample or strong caveats; for intuition, not decisions.
- Descriptive: a pattern or an association, with no causal claim.
- Projection: a forward path computed from stated assumptions about income and population, or from an assumed growth rate; it is not a tested finding and has no calibrated interval.
What this study does not establish
- A pure income effect: ticket prices are not in the models, so every income figure also includes the effect of cheaper fares. In Brazil, part of it may also be new airlines entering the market.
- An exact income figure: the United States estimate depends on how the model is set up and on how a change in airline reporting is handled. Every ranked market also fails at least one repeat of the test on pre-COVID data alone.
- A fair ranking of every market: markets count passengers differently and some have short records, so markets with close figures are treated as tied and those with no reliable estimate are left off.
- Cause and effect: the models take income as given, and the results on fares, costs and networks are associations.
- What will happen: outside Singapore the projections follow income and population only. Re-run from earlier years on the curve fitted to all years, they came in below actual growth in most pre-COVID windows, though not in all.
Licence and reuse
Ashrith Yerrapragada holds the copyright in all the writing, charts, code and analysis published on this site. All rights reserved. The code that builds the charts and the fitted models are proprietary and are not published. Full terms are on the About page.
The raw data behind this project come from the public agencies listed above and are not owned by this project. Anyone who wants to repeat an analysis should get the data directly from those agencies.
For licensing or reuse inquiries: contact@airtraffickonomics.com.