Do people fly more just because they get richer?

Across markets, air travel tracks income closely, and these charts carry the formal evidence and its limits. The models have income in them and no ticket prices, so their figures also count the fall in real fares that came with rising incomes. Charts that set other measures beside traffic, such as household budgets and income per person, sit here too. They report associations, not settled effects.

6 featured of 27 charts

Cross-market

Air travel grows roughly in step with GDP in the US, faster in Australia and far faster in Brazil

Singapore and the US are too close to rank. China, inconclusive on a shorter pre-COVID window, is shown but not ranked, and Canada is left off. Every ranked market fails at least one pre-COVID re-check of the long-run link

Field note →

airtraffickonomics.com

The ranked markets are in order of how strongly air travel moves with income. For each one, at least one repeat of the test on pre-COVID years alone disagrees about that long-run link.

descriptive Evidence strength varies by row; each row carries its own published status label.

Details

Vertical axis: Long-run GDP elasticity of passenger demand

Singapore: confirmed for existence only, because its second test gives a warning and not a pass, and that test also fails on the pre-COVID quarters alone. The United States and Brazil: confirmed, although their second test fails on the pre-COVID quarters alone. Australia: provisional, because its long-run test passes only with the pandemic quarters in the sample. China: inconclusive on its shorter window, 2011 Q1 to 2019 Q4. A status rests on the first test of a long-run link; the second only corroborates it.

Rows run from the lowest elasticity to the highest. Two neighbours are told apart only when their intervals do not overlap, the difference between them is statistically clear and no robustness check reverses them; otherwise they are tied, as Singapore and the United States are. China is shown last and not ranked. On one common window, 2005 Q1 to 2019 Q4, the same models give Singapore 1.043, the United States 1.908, Australia 1.264 and Brazil 3.169; this checks comparability and is not a second ranking (the order differs, and only Brazil passes its test there).

Singapore is a transit hub, so its elasticity blends resident demand with connecting traffic. For the United States, other specifications the data cannot reject give 0.900 to 1.315, and adjusting for a change in which carriers report gives 0.806 to 0.940. For Brazil, the carriers flying in 2000 alone give 1.12 against 2.842 for the whole market, so about 61% of that may reflect market liberalisation and not income; this rough comparison makes no causal claim. China's relation is unstable over its window, so its figure averages over a change; with a trend in the model it is not identified.

United Kingdom: its source data has no figures for 2000 to 2002, and the quarters after the gap fail the long-run test. Canada: its two traffic series disagree (passenger-kilometres give 0.984, passengers 0.192). India: a one-time drop in its passenger series leaves the long-run link unresolved. Japan: GDP barely moves over its sample, so no elasticity can be identified. Mexico: its long-run test passes, but three checks of whether the elasticity is identified go the other way. South Korea: a break in 2020 Q2 sits inside its sample, so its passing test cannot be read as a stable link.

The rows do not count the same thing: Singapore counts airport passenger movements, international only, with a transfer passenger counted twice; the United States counts enplanements (flight legs, not trips); Australia counts passengers; Brazil counts paid passengers; China counts passengers carried on Chinese airlines only. No row's model has a fare term, and real fares fell over these samples, so each elasticity is the response to income together with the fare declines that came with it. A ladder with a fare term has not been estimated.

Every whisker is a 95% interval built the same way, though the markets are estimated on samples of different lengths. Each interval holds the model's form fixed, with no time trend; where a trend is added, the estimates move and are less certain. The samples end in 2024 Q4, except that of China (2019 Q4). The ⚠ on a label marks a short sample: China (36 quarters).

How these numbers are made

Log passengers vs log GDP

Field note →

airtraffickonomics.com

Each dot is one quarter across three and a half decades. As the US economy grows, air travel grows with it, tightly and persistently.

confirmed The estimate moves with specification choices the data cannot reject.

Details

Vertical axis: log(passengers)

How these numbers are made

Cross-market

GDP helps predict air traffic in 10 of 19 markets and air traffic helps predict GDP in 5 markets

The split is GDP only in 7 markets, air traffic only in 2 markets, both in 3 markets and neither in 7 markets. Each market has its own test at the 5% level, unadjusted for the number of tests, and consumer prices, fares and oil are left out

Field note →

airtraffickonomics.com

Past income helps predict air travel in more markets than air travel helps predict income, and in many markets neither helps predict the other. These are tests of prediction only.

provisional Estimated in a small bivariate system; omitted macro drivers could alter the result.

Details

Vertical axis: Number of markets

Each market is placed by one test at one lag, and a direction counts when that test is significant at the uncorrected 5% level. The United States is counted under 'Neither' here. Its own chart, which tests every lag and uses a stricter test, reads GDP as helping to predict US air traffic.

How these numbers are made

Australian air travel grew 473% from 1985 to 2019 while GDP grew 184%

Australia: aviation demand vs GDP (growth from annual passenger and GDP sums, 1985 vs 2019, the last pre-COVID full year)

Field note →

airtraffickonomics.com

Australian air travel grows substantially faster than the economy, among the steepest responses of the ranked markets, but its long-run test passes only with the pandemic years in the sample.

provisional The long-run test behind this status passes only with the pandemic quarters in the sample; on pre-COVID data alone it fails.

Details

Window: 1985 Q1 to 2025 Q3

Vertical axis: log(passengers)

How these numbers are made

Brazilian air travel grew 230% from 2000 to 2019 while GDP grew 54%

Brazil: aviation demand vs GDP (growth from annual passenger and GDP sums, 2000 vs 2019, the last pre-COVID full year)

Field note →

airtraffickonomics.com

As a yearly rate over those years, Brazilian air travel grew almost three times as fast as the economy. Its estimated income response is the steepest of any ranked market, though part may reflect new airlines entering.

confirmed Full-sample evidence; a pre-COVID control sample dissents.

Details

Window: 2000 Q1 to 2025 Q3

Data through: 2025 Q3

Vertical axis: log(passengers)

ANAC (Brazil's civil aviation agency) scheduled flights, all carriers, domestic and international, paid passengers; non-scheduled flights are outside the passenger series plotted here.

How these numbers are made

After COVID, among the models scored on this chart, the seasonal model (SARIMAX) predicts best and the error-correction model (ECM) worst

Next-quarter forecast error (RMSE), each model refitted quarter by quarter, for the 2017 to 2019 holdout and after COVID, in ascending order of holdout error. Models see the quarter's actual macro data, and ARDL has no post-COVID score

Field note →

airtraffickonomics.com

Before the pandemic the leading models were close together. Afterwards the seasonal model predicts best and the error-correction model worst of those scored here; each sees the quarter's actual economic data.

descriptive Accuracy is conditional on feeding models the realised same-quarter macro inputs, not a real-time forecast test.

Details

Vertical axis: RMSE (lower is better)

Series: Pre-COVID holdout (2017–2019); Post-COVID generalisation (2021 Q3 onward, expanding window)

All models on this chart condition on realised same-quarter macro data: each is given the actual GDP, CPI and oil price of the quarter it forecasts, so these are not the errors of a forecast made in real time. The foundation models on the cross-family chart see macro data only up to the quarter before, so a comparison with them favours the models shown here. The expanding-window ElasticNet predictions are capped at three standard deviations from the training mean.

ARDL has no post-COVID score on this chart, so the post-COVID ranking covers the other models only. The ARDL plotted here is a forecasting version fitted on data to the end of 2019. It is a different model from the one behind the site's headline elasticity, and that model is not scored on this chart.

The 'static' column mixes four ways of forecasting that cannot be compared, so its ranking of the models is not like for like. The expanding-window column is the matched comparison.

How these numbers are made

Every chart on this question

BR Brazil: passenger decomposition, incumbent vs new entrant
Brazil's passengers are split between airlines that were already flying before GOL launched and those that entered since. It is an informal comparison by carrier and does not say what caused the growth.
US Vehicles absorb about 5% of household income year after year; flying never tops 0.7%
After vehicle purchases, gasoline is the next largest share of income and airline fares the third. Small as it is, the airfare share is several times that of mass transit and intercity trains combined.
CN Chinese passenger traffic grew 127% from 2011 to 2019 while GDP grew 74%
Each dot plots one quarter's Chinese passengers against economic output. Before the pandemic the dots climb in a tight line; the pandemic quarters fall far below it, and the latest sit under where that line would lead.
US Intervals promising 90% coverage delivered as little as 17% across the COVID break
Standard forecast uncertainty bands broke down across the COVID shock. An adaptive method brought coverage back toward its promise, though still short of it, for the models it covers.
Cross-Market Australia's elasticity does not depend on any one crisis control; Brazil's and Singapore's do
Australia, Brazil and Singapore still pass the bounds test and stay inside their published intervals when any control but the pandemic-trough one is dropped. Without that one, Canada's off-ladder estimate roughly halves.
US After COVID, the three foundation models forecast with a lower error than the five traditional models scored
In the COVID quarters every model's error is several times its error afterwards, and two of the three foundation models go from lowest on the chart to highest. Foundation models say nothing about why demand moves.
US Nevada and Hawaii fly far more than their GDP predicts; Delaware flies least
Some states fly far more than their GDP alone would give: Nevada, Hawaii and Florida top the list, and Delaware sits at the bottom.
US Household airline spending
The average household's airline spending has roughly doubled in dollars since the early twenty-tens, with a deep pandemic dip on the way. As a share of income it has risen much less.
US Household transport spending by mode
Vehicle purchases take the overwhelming share of household spending on the modes shown, and air travel around a tenth. Fuel and other running costs are left out, so air's share of all transport spending is smaller.
US Real GDP and potential GDP
US output has more than doubled in real terms over three and a half decades and now runs slightly above its estimated potential.
US Real disposable income vs air travel intensity
As real income per person rose over three and a half decades, flights per person rose by about two-thirds, and the two moved together over the long run.
US State GDP per capita vs passengers per capita
Each dot is a state, placed by income per person and flights per person. Some states fly many times more than similarly rich peers.
EU Air passengers in 14 EU countries grew 115% from 2002 to 2019 while their real GDP grew 23%
The aggregate sums the EU countries that report in every quarter, and its income slope depends on which countries are summed. Only one of the two tests for a long-run link finds one, so the slope is a description.
EU Cumulative GDP vs passenger growth (2002–2019)
Across the pre-COVID window, passengers grew roughly tenfold in Latvia, Romania, Slovakia, Poland and Lithuania, the highest dots on the chart. Four of the five were also among the fastest-growing economies shown.
IN Indian domestic air travel grew 178% from 2010 to 2019 while GDP grew 79%
Indian domestic passengers against economic output, one dot per quarter. The dots climb together before the pandemic and fall far below in it; whether they share a lasting link is unresolved, so nothing is fitted.
KR Passengers through Korean airports grew 144% from 2005 to 2019 while GDP grew 63%
Quarter by quarter, passengers at Korean airports plotted against economic output. The dots climb together until the pandemic and drop far below in it; nothing is fitted, and a domestic trip is counted at both ends.
MX Mexican air travel grew 160% from 2000 to 2019 while GDP grew 36%
Before the pandemic Mexican air travel grew far faster than the economy. No income figure is published for Mexico because its estimate does not survive the checks applied to it.
Cross-Market When air traffic and GDP drift apart, it is air traffic that closes the gap
In the few markets where a long-run link is supported, when air traffic and national income drift apart it is air traffic that adjusts back; income barely moves. Most of the markets tested are not on this chart.
SG Singaporean air travel grew about 10-fold since 1980 while GDP grew 13-fold
More than four decades of Singapore data show income and air travel climbing together over a very wide range.
UK UK air travel grew 94% from 1998 to 2019 while GDP grew 48%
Each dot plots one quarter's UK passengers against economic output. The dots rise together in a broad band before the pandemic and drop far below it in the pandemic quarters; nothing is fitted.
US GDP helps predict US air travel at 2 of 8 lags tested, while air travel helps predict GDP at no lag
In the United States, past income helps predict air travel at some lags and past air travel never helps predict income; a stricter version of the test agrees. Both are tests of prediction only.