FlyerIntel Research
Who Cancels More When the Weather Turns
Each airline’s bad-weather cancellations, set against what the other airlines did at the same airports in the same months.
Published October 12, 2026 · Source period July 2023 – July 2026
Summary
United Airlines cancelled 2.9% of its bad-weather departures against 5.0% expected from the other airlines at the same airports and months. PSA Airlines cancelled 10.9% of its bad-weather departures against 6.2% expected from the other airlines at the same airports and months. Of 13 airlines, 4 cancelled fewer bad-weather departures than the others at the same airports and 6 cancelled more. Envoy Air and SkyWest Airlines cancelled fewer than the others in normal weather and more in bad weather.
Key findings
- 1,428,157 bad-weather departures from 13 airlines were compared with the other airlines’ rates at the same airports and in the same months.
- United Airlines cancelled 2.9% of its bad-weather departures against 5.0% expected from the other airlines at the same airports and months, 0.6 times the expected number (1.0% against 1.3% in normal weather).
- PSA Airlines cancelled 10.9% of its bad-weather departures against 6.2% expected from the other airlines at the same airports and months, 1.8 times the expected number (2.6% against 1.4% in normal weather).
- Envoy Air cancelled 1.1% against 1.4% in normal weather and 8.8% against 7.2% in bad weather.
- Southwest Airlines cancelled 0.5 times the expected number in normal weather and 0.8 times in bad weather, a smaller edge over the other airlines when the weather turned.
- United Airlines cancelled 0.6 times the expected number in bad weather against 0.8 times in normal weather.
- At its 3 busiest airports Alaska Airlines cancelled 1.5 times the expected number in bad weather, against 0.8 times elsewhere.
What was compared
A raw cancellation rate mixes where an airline flies with how it runs. An airline that flies mostly into stormy airports will cancel more than one that does not, whatever its habits. So each airline’s bad-weather departures were compared with what the other airlines did at the same airport, in the same calendar month of the same year, in hours with the same kind of weather. Each such combination is a stratum. For every stratum the others’ cancellation rate was applied to the airline’s own departures, and the results were added up. That sum is the number of cancellations the airline would have had if it had done what the others did. 1,428,157 bad-weather departures from 13 airlines went into the comparison.
Cancellations in bad weather and in normal weather, against the other airlines at the same airports
| Airline | Bad-weather departures compared | Cancelled | Others, same airports and months | Ratio | Against others | Normal weather: cancelled | Others | Ratio | Against others |
|---|---|---|---|---|---|---|---|---|---|
| United Airlines | 200,900 | 2.9% | 5.0% | 0.57 | Fewer | 1.0% | 1.3% | 0.79 | Fewer |
| Delta Air Lines | 261,713 | 2.7% | 3.9% | 0.69 | Fewer | 1.0% | 1.3% | 0.75 | Fewer |
| Southwest Airlines | 215,674 | 2.8% | 3.7% | 0.77 | Fewer | 0.7% | 1.3% | 0.51 | Fewer |
| American Airlines | 216,274 | 5.5% | 6.1% | 0.90 | Fewer | 1.4% | 1.3% | 1.04 | Level |
| Alaska Airlines | 48,652 | 2.4% | 2.4% | 1.00 | Level | 1.1% | 0.9% | 1.24 | More |
| JetBlue | 61,502 | 5.2% | 5.1% | 1.01 | Level | 1.8% | 1.4% | 1.25 | More |
| Frontier Airlines | 45,005 | 4.6% | 4.3% | 1.07 | Level | 1.8% | 1.1% | 1.67 | More |
| Spirit Airlines | 47,575 | 5.1% | 4.2% | 1.22 | More | 1.7% | 1.2% | 1.46 | More |
| Envoy Air | 44,539 | 8.8% | 7.2% | 1.23 | More | 1.1% | 1.4% | 0.76 | Fewer |
| Republic Airways | 98,233 | 8.7% | 5.7% | 1.53 | More | 2.4% | 1.6% | 1.51 | More |
| Endeavor Air | 21,435 | 5.2% | 3.3% | 1.59 | More | 2.0% | 1.2% | 1.65 | More |
| SkyWest Airlines | 126,591 | 5.4% | 3.1% | 1.75 | More | 0.8% | 0.9% | 0.80 | Fewer |
| PSA Airlines | 40,064 | 10.9% | 6.2% | 1.76 | More | 2.6% | 1.4% | 1.83 | More |
Departures in the weather hour of the scheduled departure, at covered airports. Ratio is the airline’s cancellations divided by the number the other airlines’ rates would give its departures, stratum by stratum. Fewer or More needs a gap of 3 standard errors and a ratio at least 10% from one. Airlines with fewer than 10,000 matched departures in either kind of weather are not shown.
Who cancels fewer, and who cancels more
United Airlines, Delta Air Lines, Southwest Airlines, and American Airlines cancelled fewer bad-weather departures than the other airlines at the same airports, Spirit Airlines, Envoy Air, Republic Airways, Endeavor Air, SkyWest Airlines, and PSA Airlines cancelled more, and Alaska Airlines, JetBlue, and Frontier Airlines were level. United Airlines cancelled 2.9% of its bad-weather departures against 5.0% expected from the other airlines at the same airports and months, 0.6 times the expected number. PSA Airlines cancelled 10.9% of its bad-weather departures against 6.2% expected from the other airlines at the same airports and months, 1.8 times the expected number.
The bad-weather record is not the everyday record
Envoy Air and SkyWest Airlines cancelled fewer flights than the other airlines in normal weather and more in bad weather. Envoy Air cancelled 1.1% against 1.4% in normal weather and 8.8% against 7.2% in bad weather. Southwest Airlines cancelled fewer than the others in both kinds of weather, and the edge was smaller in bad weather (Southwest Airlines: ratio 0.51 in normal weather, 0.77 in bad weather). United Airlines did better against the others in bad weather than in normal weather, so the gap widened when the weather turned (United Airlines: ratio 0.79 in normal weather, 0.57 in bad weather). Spirit Airlines, Republic Airways, Endeavor Air, and PSA Airlines cancelled more than the others in both kinds of weather, so for those airlines bad weather is not the only place the gap shows. The rate on an ordinary day is a poor guide to how an airline handles a storm.
At home and elsewhere: bad-weather cancellations at an airline’s 3 busiest airports
| Airline | Its 3 busiest airports | Departures there | Cancelled | Others | Ratio | Elsewhere: cancelled | Others | Ratio |
|---|---|---|---|---|---|---|---|---|
| Alaska Airlines | SEA, PDX, ANC | 23,906 | 1.9% | 1.2% | 1.49 | 2.9% | 3.4% | 0.83 |
| SkyWest Airlines | DEN, ORD, LAX | 51,488 | 6.8% | 3.3% | 2.07 | 4.5% | 3.0% | 1.51 |
| Spirit Airlines | FLL, MCO, LAS | 8,273 | 5.2% | 3.2% | 1.63 | 5.0% | 4.4% | 1.15 |
| Endeavor Air | LGA, ATL, JFK | 12,581 | 6.7% | 4.0% | 1.67 | 3.0% | 2.2% | 1.40 |
| Southwest Airlines | DEN, LAS, BWI | 54,928 | 3.4% | 3.6% | 0.93 | 2.6% | 3.7% | 0.71 |
| Delta Air Lines | ATL, MSP, DTW | 122,148 | 2.8% | 3.4% | 0.81 | 2.7% | 4.4% | 0.61 |
| Frontier Airlines | DEN, MCO, ATL | 15,508 | 4.1% | 3.4% | 1.20 | 4.8% | 4.8% | 1.02 |
| JetBlue | BOS, JFK, FLL | 34,968 | 5.6% | 5.4% | 1.03 | 4.6% | 4.7% | 0.99 |
| American Airlines | DFW, CLT, MIA | 78,450 | 7.6% | 8.4% | 0.91 | 4.2% | 4.8% | 0.89 |
| Envoy Air | DFW, ORD, MIA | 31,255 | 10.1% | 8.2% | 1.22 | 5.9% | 4.7% | 1.25 |
| United Airlines | DEN, ORD, IAH | 78,601 | 2.7% | 5.4% | 0.50 | 3.0% | 4.7% | 0.63 |
| Republic Airways | LGA, BOS, DCA | 38,494 | 10.6% | 7.3% | 1.45 | 7.4% | 4.6% | 1.61 |
| PSA Airlines | CLT, DCA, DFW | 27,765 | 11.2% | 6.7% | 1.67 | 10.0% | 4.9% | 2.06 |
Bad-weather departures only. An airline’s busiest airports are the 3 where it has the most scheduled departures in the data. Airlines with fewer than 5,000 matched departures at either its busiest airports or elsewhere are not shown.
Does home help
Alaska Airlines cancelled 1.5 times the expected number at its busiest airports (SEA, PDX, ANC) and 0.8 times elsewhere. PSA Airlines was the other way: 1.7 times the expected number at CLT, DCA, DFW and 2.1 times elsewhere. At an airport where an airline has most of the departures, the expected number is built from the few departures of the others.
What this does not show
Fewer cancellations is not the same as a better day for passengers. An airline that keeps flying through a storm can land its flights late, and one that cancels early can get the rest in on time. The companion report Cancel Early or Run Late takes the arrivals of the flights that did operate. The comparison holds the airport and the month, not the hour of the day, so an airline that schedules more of its bad-weather departures into the afternoon thunderstorm peak is held to the same standard as one that does not. A regional carrier flying for a mainline brand is compared under its own name, and the mainline brand’s schedule is part of what it is asked to fly. 117,550 bad-weather departures were left out because the other airlines had fewer than 120 departures, or fewer than 120 completed flights, in the stratum.
Methodology
Calculated by FlyerIntel from 21,708,321 individual flight records in the US Department of Transportation Bureau of Transportation Statistics Reporting Carrier On-Time Performance dataset, covering July 2023 through July 2026. Following BTS methodology, a flight counts as delayed when it arrives 15 or more minutes after its scheduled arrival time. Cancelled and diverted flights have no arrival time and are excluded from on-time rates rather than counted as late; cancellation rates are reported separately. Full definitions are published at /methodology/. Weather comes from the hourly observation at the airport, matched to the scheduled departure hour in the airport’s own time zone. Bad weather means IFR or LIFR conditions, or a thunderstorm, snow or ice, fog, heavy rain or high wind; normal weather means VFR with none of those. The figures are read from the airline-by-airport-by-month weather table, which counts scheduled departures. An airline’s expected cancellations are its departures in each stratum (one airport, one month of one year, one kind of weather) multiplied by the other airlines’ cancellation rate in that stratum; a stratum counts only when the others have at least 120 departures and 120 completed flights in it. A standard error comes from the binomial variance of the expected count, and Fewer or More is printed only for a gap of 3 standard errors and a ratio at least 10% from one.
Limitations
The source dataset covers scheduled domestic flights operated by carriers above the DOT reporting threshold; smaller carriers and international flights are not included. Flights sold under a mainline brand but flown by a regional carrier are attributed to the operating carrier, which is how the DOT records them. Past performance describes what has happened and is not a prediction about any individual flight. The standard errors treat flights as independent, and a storm that cancels a whole bank is not, so they are smaller than the real uncertainty; the verdict thresholds are a screen and not a test. The comparison cannot see why a flight was cancelled, whether the airline decided early, or what it did for the passengers.
Sources
- FlyerIntel analysis — FlyerIntel · source
- Iowa Environmental Mesonet ASOS/METAR archive — Iowa Environmental Mesonet, Iowa State University · source
- Reporting Carrier On-Time Performance — US Department of Transportation, Bureau of Transportation Statistics · source
Full FlyerIntel methodology · Report a problem with this analysis