A 10% longer flight costs about 3% more, so fares rise far more slowly than distance
View data table
Each dot is a route, sized by its passengers, and the line is the fit to the dots. Among short and medium routes, fares at a given distance still differ several-fold.
descriptive A pattern in the data, not a causal estimate.
Details
Vertical axis: Average fare ($)
Series: Routes; Log-log fit to the plotted routes
On the full 2020 to 2025 route-quarter panel the distance elasticity is 0.23 in both of its eras (2020 and 2021, and 2022 on), flatter than the 0.33 on the plotted routes. The panel starts in 2020, so these two eras compare the pandemic years with the recovery, not the years before and after COVID.
Longer flights cost more in total but far less per mile
View data table
The shortest flights cost roughly six times as much per mile as the longest, even though the longest cost far more in total.
Details
Vertical axis: Average fare ($)
Series: Average fare ($); Fare per mile (¢)
Fares are higher on longer routes and where more travel is on unrestricted or high-fare ticket codes, and lower where more trips stay on one airline
View data table
Each bar is how the fare moves with one route attribute when the others are held constant. Fares in this fit are also lower on busier routes and on more concentrated ones.
descriptive A pattern in the data, not a causal estimate.
Details
Vertical axis: Coefficient (log-linear)
The bars are in a fixed order and are not ranked by size.
The "Unrestricted-fare-basis share (F/C/D)" regressor is the share of coupons on unrestricted or high-fare fare bases, a code each carrier defines. It stands in for fare restrictions, and its positive coefficient is an association with those restrictions, not a price premium for the first cabin.
Long-haul routes are fewer, not thinner
View data table
The longer distance bands hold fewer routes, but passengers per route are broadly similar in every band, with the longer bands slightly ahead.
Details
Vertical axis: Passengers / count
Series: Mean annual passengers; Number of routes
The bars are raw means of the routes in each of the four distance bands, with no dispersion or interval, and the band means are not monotone (the third band sits below the second).
Average fares in nominal dollars now sit above their 2020 Q1 level
View data table
Before adjusting for inflation, average US fares sank during COVID, peaked during the reopening and now sit above their level at the start of the pandemic.
Details
Window: 2020 Q1 to 2025 Q2
Data through: 2025 Q2
Vertical axis: Average fare (nominal $)
Series: Average fare (nominal $); Median fare (nominal $); 25th percentile fare (nominal $); 75th percentile fare (nominal $)
Latest: Average fare (nominal $) 240.5; Median fare (nominal $) 203; 25th percentile fare (nominal $) 130; 75th percentile fare (nominal $) 299.5
Within each distance band from 500 to 2,000 miles, routes with more carriers have higher fares per mile
View data table
The chart keeps only routes with enough sampled tickets. Among them, more airlines go with higher fares per mile in most distance bands; only distance is held fixed, so this cannot be read as the effect of competition.
descriptive A pattern in the data, not a causal estimate.
Details
Vertical axis: Fare per mile ($)
Series: Monopoly; Duopoly; 3+ carriers
In 2024 Q3, routes with one carrier average 603 sampled passengers a route against 1,157 on routes with three or more carriers. The comparison holds the distance band fixed and nothing else. It does not hold fixed how busy a route is, which airlines fly it or how many of its passengers connect, and it covers one quarter.
Each bar is the passenger-weighted mean fare per mile of the routes in one distance band and one carrier-count group. A route is a directional airport pair in the DB1B ticket sample, connecting itineraries included, with at least 100 sampled passengers in the quarter; a route with one carrier is one on which a single airline sold all the sampled tickets, not a legal monopoly.
Distance still deters flying: 1% farther means roughly 0.49% fewer passengers
View data table
The first bar is the headline estimate: traffic thins as distance grows. The other bars are reference models with known weaknesses, shown for comparison and not as independent confirmation.
confirmed
Details
Vertical axis: Distance elasticity (log-log)
The bars are in a fixed order and are not ranked by size.
The Heckman bar uses the same network controls as the augmented model, which are outcomes of demand, so read it as conditional on them. It is also unidentified: the variable its correction relies on (whether the two airports share a state) fails its relevance test (p=0.48), and the interval of its selection term (the Mills ratio) crosses zero. It is a directional robustness check only.