Worth noting that the sharp increase in schedule padding that emerges around 2009 is an artifact of regulatory changes. In 2009, the DOT finalized a rule against "unrealistic scheduling," as well as specifically forbidding "chronically delayed flights": https://www.ecfr.gov/current/title-14/chapter-II/subchapter-F/part-399
Since chronically delayed flights are defined as >30 min past scheduled arrival times, it caused airlines to start padding their schedules. "Unrealistic" scheduling isn't clearly defined, but that probably encourages even more conservatism in arrival estimates.
- more flights are cancelled today when there is a question of aircraft reliability
- fewer flights are cancelled today when there would be a long delay*
That hypothesis could get you more very long delays, fewer safety accidents, and roughly constant cancellation rates. Maybe it also helps explain why we do not observe increases in minor delays after 1997 (if minor delays didn’t usually result in cancellation in the past), which I think we would expect from the congestion hypothesis.
*why would that be? Maybe to offset increased safety related cancellations? Maybe because airports are nicer and customers are more willing to wait in them? Not sure if that’s true. Or maybe it’s cheaper than booking everyone hotels? Or maybe it’s a consequence of greater logistical sophistication enabling airlines to avoid cascading problems from allowing flights to run very late?
Hmm that's interesting. Maybe one reason airlines are keeping more delayed flights instead of cancelling them is because each flight has a lot more people on average, so the compensation and rebooking would be more difficult.
I do also agree that we should expect a rise in both shirt and long delays from the congestion hypothesis, but congestion + schedule padding can explain rising long delays and falling/stable short delays.
I have noticed this same phenomenon and it's an excellent thing. The problem with a delayed flight is not the wait to depart, it's the missing of scheduled events, such as connecting flights, by arriving late. By padding their schedules they provide travellers with much lower-risk timetables, and arriving early is practically never an inconvenience. This is not perfidy, this is sensible timetabling.
Very helpful data-grounded discussion piece. 😍 To add an airliner perspective on why flight times are scheduled to be longer: Airports have become significantly larger and more sprawling. Taxiing (the time between gate and runway, either on arrival or departure) probably takes a few minutes longer. Of course this is partially due to "congestion" (a "bad" thing), but there is also an element of infrastructure just getting more developed (a "good" thing). Also, flights spend more time in holding patterns, and airlines plan for that. E.g., scheduled flight times from A to B may be 10min longer than the return flight from B to A. Holding patterns probably fall completely under the "congestion" header, as this is a function mainly of having sufficiently runways and airspace around the airport.
As a former airline pilot I challenge your claims. I suspect that, if anything, there is far less holding now than there used to be. This is because the airlines tend to carry minimum fuel (to save on weight, and thus fuel) and when things go wrong, they divert a lot more quickly than they used to. This is net worse for everyone than holding, but is less holding, not more.
I'm not sure about larger and more sprawling airports--this is true in some places (the new runway at ATL for instance), but the original article indicates that there's been almost no new runway construction.
I am not familiar with domestic US airline ops, only international airlines operating into US. But can confirm that generally, how much fuel gets taken is more monitored (bc you burn around 3% of extra fuel per hour, which impacts costs). Probably aircraft carry less fuel than in 1990, adjusted by stage length. But I am not sure whether there is a data source to confirm this - and whether the impact of diversions is material to affect delays. Let's keep in mind that diversions are awfully expensive from a financial perspective (extra fuel for extra leg, extra airport fees, affecting crew rosters), so I am sure no airline is happy to have so many diversions that these extra costs overcompensate the fuel savings.
As for airport size - agree, not too many runways were added across the US, but terminals were either remodeled, extended or built from scratch. Just take JFK - T1 in 1998, T4 2001, T5 in 2008, T8 in 2007. Not all of these were totally new, but you added new taxiways, aprons were remodeled, etc. This adds to taxiing times. 5min more is already material.
It's just that the base rate of 2 is still quite low; one or two percent of flights.
So the severe delays can go up by 4x, but if short delays go down by 30% because of the schedule padding, that can totally mask the number of 15+ minute delays because the vast majority are the short delays that are going down in frequency.
15+ minute delays have remained stable, but 1+, 1.5+, and 3hr+ delays have all grown, which means that 15-45 minute delays must have declined, since 15+ is the sum of all the delay lengths.
Just came from 3 weeks in Bali. Both airports in Indonesia were beautiful, efficient and a pleasure to arrive and depart from. Unfortunately we left from LAX and returned to LAX. LAX was perhaps the low point of the trip in both directions. Dim and dingy. Long lines that did not exist in the other places. No signage directing passengers which way to go. I was so disappointed. The amount of tax money that is spent on continuous remodeling is going into a hole somewhere. Another example of bad California politics.
Not sure it's a CA thing. JFK is very similar to your description of LAX, whereas SFO is world-class, beautiful and efficient and comfortable despite being run by the city government of SF.
I’d bet there’s a trade off between very long delays and cancellations. (I also think there’s been some regulatory changes affecting how airlines think about delay vs cancellation.) What’s the trend look like for the combined rate of cancellations/very long delays?
Well delays are growing a lot and cancellations are basically staying stable on average or even getting slightly more common, so the combined trend is still going to be increasing
One thing that isn't considered in your excellent analysis...and perhaps cannot be, since I doubt there's a readily accessible data set for it: the actual time it takes to complete a trip by air, taking into account the latency involved at both ends.
Air travel isn't just about the time you spend on the plane. Indeed I would concur that the actual time spent taxiing and flying isn't any worse than it used to be. What HAS gotten worse--at least in my experience, which goes back to when I was following my parents onto airplanes as a pup in the 1960s--is the airport experience at both ends.
I suspect that if you did point-to-point (i.e. how long does it take from when you arrive at the departure airport terminal to when you leave the destination airport terminal) it would look a lot worse, owing to such factors as increased security and the massive expansion of the airports themselves.
I spent nearly forty years (1981-2020) in the National Capital Region and I saw massive expansions of all three airports, which contributed mightily to both the time and discomfort of the experience. Terminals are a mob scene; the monorails that are supposed to effortlessly whisk you from point A to point B seem to have been designed by the Armour Hot Dog Company; and waiting areas at the gates are SRO.
Thanks to the hub-and-spoke architecture of modern airports, it now takes much longer to get from "curbside" to "planeside," and that's just the beginning: O'Hare and Atlanta, for example, have immense terminals that take eons to traverse, let alone--God forbid!--if you have to change from one to another.
Even much more modest terminals suffer from this: I flew into Phoenix Sky Harbor last year, for the first time since 1994 (!), and it took me a good fifteen minutes to get from my arrival gate to what turned out to be the monorail terminal (!), which then took another fifteen minutes to take me to the car-rental area. By the time I actually got off the airport campus it was nearly an hour since I'd "arrived."
So while I accept your analysis as valid, I think it fails to take into account the costs imposed on travelers by the airlines and the airport authorities while on the ground. But--again--you are not at fault here, so please do not take this as criticism.
- "The amount of congestion from the same number of larger planes isn’t as obvious or intense as congestion from a larger fleet, but larger aircraft concentrate passenger flows into the same peak-hour runway slots, terminal gates, and baggage systems, ..."
Aren't these second-order effects, whereas the first-order effect is to actually increase capacity without needing new runways, terminals, and redesigned airspace? Obviously those things would be great, but are hard to do. In the face of those constraints, upgauging is something we can do today!
- "I don’t have detailed data on air travel in Canada or New Zealand but their delay rates don’t seem obviously better than in the US"
Agree it's hard to make exact comparisons, but New Zealand seems perhaps 5-10% better, whether you look at A0 or A15 to account for block time padding. In Canada, it'd be great to separate out de-icing delays, which are common and idiosyncratic to Canada, and just look at ATC delays.
An interesting breakdown! And further interesting discussion here in the comments.
Is there any way to measure root cause of the delay? Various forms of congestion, ATC shortages, etc are undoubtedly part of the increase, but I also wonder how much of this is due to increasing weather events delaying or canceling flights. My anecdotal experience with delays as a traveler is that “weather” is the most cited cause, and given the documented incline in severe weather globally I wonder if this is in fact putting measurable stress on the flight network.
An interesting breakdown! And further interesting discussion here in the comments.
Is there any way to measure root cause of the delay? Various forms of congestion, ATC shortages, etc are undoubtedly part of the increase, but I also wonder how much of this is due to increasing weather events delaying or canceling flights. My anecdotal experience with delays as a traveler is that “weather” is the most cited cause, and given the documented incline in severe weather globally I wonder if this is in fact putting measurable stress on the flight network.
I've no idea if it's still true, but there was a phase during which airlines would push back from the terminal, trundle to a holding area, then park the plane and power it down. The reason that was given was that they wanted to clear the gate for the next user, notwithstanding the lack of a takeoff slot, but many folks surmised that the hidden agenda was to artificially bump up the "on-time departure" stats.
Again, I do not know if these practices are still in effect, but when I was traveling from Dulles to Heathrow on a regular basis in the early part of the century, it was...pretty much the routine. Perhaps you could comment on this.
> The connection between this financialization of air travel to increasing delays is difficult to make explicit, but something about the coincidence of credit-card-mile loans shifting transportation into a loss-leader sector, airfare prices falling, and a rapid increase in the rate of long delays is too suggestive to ignore.
The obvious question is, if the airlines make so much money on credit cards but lose so much on air transport, why not just stop doing the air transport? And of course, the answer is "nobody will use your credit card if your air transport offering sucks". In-cabin quality (outside basic economy) has improved substantially in the past two decades, especially on Delta and United; they need to keep investing in this to keep the miles / credit card dollars flowing, so I'm not sure it's fully accurate to treat these as two independent lines of business.
Worth noting that the sharp increase in schedule padding that emerges around 2009 is an artifact of regulatory changes. In 2009, the DOT finalized a rule against "unrealistic scheduling," as well as specifically forbidding "chronically delayed flights": https://www.ecfr.gov/current/title-14/chapter-II/subchapter-F/part-399
Since chronically delayed flights are defined as >30 min past scheduled arrival times, it caused airlines to start padding their schedules. "Unrealistic" scheduling isn't clearly defined, but that probably encourages even more conservatism in arrival estimates.
Nice piece! An alternative hypothesis:
- more flights are cancelled today when there is a question of aircraft reliability
- fewer flights are cancelled today when there would be a long delay*
That hypothesis could get you more very long delays, fewer safety accidents, and roughly constant cancellation rates. Maybe it also helps explain why we do not observe increases in minor delays after 1997 (if minor delays didn’t usually result in cancellation in the past), which I think we would expect from the congestion hypothesis.
*why would that be? Maybe to offset increased safety related cancellations? Maybe because airports are nicer and customers are more willing to wait in them? Not sure if that’s true. Or maybe it’s cheaper than booking everyone hotels? Or maybe it’s a consequence of greater logistical sophistication enabling airlines to avoid cascading problems from allowing flights to run very late?
Hmm that's interesting. Maybe one reason airlines are keeping more delayed flights instead of cancelling them is because each flight has a lot more people on average, so the compensation and rebooking would be more difficult.
I do also agree that we should expect a rise in both shirt and long delays from the congestion hypothesis, but congestion + schedule padding can explain rising long delays and falling/stable short delays.
I have noticed this same phenomenon and it's an excellent thing. The problem with a delayed flight is not the wait to depart, it's the missing of scheduled events, such as connecting flights, by arriving late. By padding their schedules they provide travellers with much lower-risk timetables, and arriving early is practically never an inconvenience. This is not perfidy, this is sensible timetabling.
Very helpful data-grounded discussion piece. 😍 To add an airliner perspective on why flight times are scheduled to be longer: Airports have become significantly larger and more sprawling. Taxiing (the time between gate and runway, either on arrival or departure) probably takes a few minutes longer. Of course this is partially due to "congestion" (a "bad" thing), but there is also an element of infrastructure just getting more developed (a "good" thing). Also, flights spend more time in holding patterns, and airlines plan for that. E.g., scheduled flight times from A to B may be 10min longer than the return flight from B to A. Holding patterns probably fall completely under the "congestion" header, as this is a function mainly of having sufficiently runways and airspace around the airport.
As a former airline pilot I challenge your claims. I suspect that, if anything, there is far less holding now than there used to be. This is because the airlines tend to carry minimum fuel (to save on weight, and thus fuel) and when things go wrong, they divert a lot more quickly than they used to. This is net worse for everyone than holding, but is less holding, not more.
I'm not sure about larger and more sprawling airports--this is true in some places (the new runway at ATL for instance), but the original article indicates that there's been almost no new runway construction.
I am not familiar with domestic US airline ops, only international airlines operating into US. But can confirm that generally, how much fuel gets taken is more monitored (bc you burn around 3% of extra fuel per hour, which impacts costs). Probably aircraft carry less fuel than in 1990, adjusted by stage length. But I am not sure whether there is a data source to confirm this - and whether the impact of diversions is material to affect delays. Let's keep in mind that diversions are awfully expensive from a financial perspective (extra fuel for extra leg, extra airport fees, affecting crew rosters), so I am sure no airline is happy to have so many diversions that these extra costs overcompensate the fuel savings.
As for airport size - agree, not too many runways were added across the US, but terminals were either remodeled, extended or built from scratch. Just take JFK - T1 in 1998, T4 2001, T5 in 2008, T8 in 2007. Not all of these were totally new, but you added new taxiways, aprons were remodeled, etc. This adds to taxiing times. 5min more is already material.
I can't wrap my head around two things you show:
1. Flights are taking longer now than in 1990, but airlines are scheduling them to be substantially longer.
2. There is a sharp rise in severe delays, as measured by actual vs scheduled without corrections
1 acts to decrease the reported delay figures, 2 acts to increase it - so my sleepy brain is struggling to fit both together
It's just that the base rate of 2 is still quite low; one or two percent of flights.
So the severe delays can go up by 4x, but if short delays go down by 30% because of the schedule padding, that can totally mask the number of 15+ minute delays because the vast majority are the short delays that are going down in frequency.
But you show that short delays have remained stable?
15+ minute delays have remained stable, but 1+, 1.5+, and 3hr+ delays have all grown, which means that 15-45 minute delays must have declined, since 15+ is the sum of all the delay lengths.
Got it, I thought they were intervals like 15 mins - 1 hr, 1hr - 1.5hr, etc
Just came from 3 weeks in Bali. Both airports in Indonesia were beautiful, efficient and a pleasure to arrive and depart from. Unfortunately we left from LAX and returned to LAX. LAX was perhaps the low point of the trip in both directions. Dim and dingy. Long lines that did not exist in the other places. No signage directing passengers which way to go. I was so disappointed. The amount of tax money that is spent on continuous remodeling is going into a hole somewhere. Another example of bad California politics.
Not sure it's a CA thing. JFK is very similar to your description of LAX, whereas SFO is world-class, beautiful and efficient and comfortable despite being run by the city government of SF.
I’d bet there’s a trade off between very long delays and cancellations. (I also think there’s been some regulatory changes affecting how airlines think about delay vs cancellation.) What’s the trend look like for the combined rate of cancellations/very long delays?
Well delays are growing a lot and cancellations are basically staying stable on average or even getting slightly more common, so the combined trend is still going to be increasing
One thing that isn't considered in your excellent analysis...and perhaps cannot be, since I doubt there's a readily accessible data set for it: the actual time it takes to complete a trip by air, taking into account the latency involved at both ends.
Air travel isn't just about the time you spend on the plane. Indeed I would concur that the actual time spent taxiing and flying isn't any worse than it used to be. What HAS gotten worse--at least in my experience, which goes back to when I was following my parents onto airplanes as a pup in the 1960s--is the airport experience at both ends.
I suspect that if you did point-to-point (i.e. how long does it take from when you arrive at the departure airport terminal to when you leave the destination airport terminal) it would look a lot worse, owing to such factors as increased security and the massive expansion of the airports themselves.
I spent nearly forty years (1981-2020) in the National Capital Region and I saw massive expansions of all three airports, which contributed mightily to both the time and discomfort of the experience. Terminals are a mob scene; the monorails that are supposed to effortlessly whisk you from point A to point B seem to have been designed by the Armour Hot Dog Company; and waiting areas at the gates are SRO.
Thanks to the hub-and-spoke architecture of modern airports, it now takes much longer to get from "curbside" to "planeside," and that's just the beginning: O'Hare and Atlanta, for example, have immense terminals that take eons to traverse, let alone--God forbid!--if you have to change from one to another.
Even much more modest terminals suffer from this: I flew into Phoenix Sky Harbor last year, for the first time since 1994 (!), and it took me a good fifteen minutes to get from my arrival gate to what turned out to be the monorail terminal (!), which then took another fifteen minutes to take me to the car-rental area. By the time I actually got off the airport campus it was nearly an hour since I'd "arrived."
So while I accept your analysis as valid, I think it fails to take into account the costs imposed on travelers by the airlines and the airport authorities while on the ground. But--again--you are not at fault here, so please do not take this as criticism.
Where do you post your data? I'd love to re-create some pieces of your analysis
Thanks for writing - really enjoyed this!
Two comments / questions:
- "The amount of congestion from the same number of larger planes isn’t as obvious or intense as congestion from a larger fleet, but larger aircraft concentrate passenger flows into the same peak-hour runway slots, terminal gates, and baggage systems, ..."
Aren't these second-order effects, whereas the first-order effect is to actually increase capacity without needing new runways, terminals, and redesigned airspace? Obviously those things would be great, but are hard to do. In the face of those constraints, upgauging is something we can do today!
- "I don’t have detailed data on air travel in Canada or New Zealand but their delay rates don’t seem obviously better than in the US"
Agree it's hard to make exact comparisons, but New Zealand seems perhaps 5-10% better, whether you look at A0 or A15 to account for block time padding. In Canada, it'd be great to separate out de-icing delays, which are common and idiosyncratic to Canada, and just look at ATC delays.
https://www.transport.govt.nz/area-of-interest/air-transport/airline-on-time-performance
An interesting breakdown! And further interesting discussion here in the comments.
Is there any way to measure root cause of the delay? Various forms of congestion, ATC shortages, etc are undoubtedly part of the increase, but I also wonder how much of this is due to increasing weather events delaying or canceling flights. My anecdotal experience with delays as a traveler is that “weather” is the most cited cause, and given the documented incline in severe weather globally I wonder if this is in fact putting measurable stress on the flight network.
An interesting breakdown! And further interesting discussion here in the comments.
Is there any way to measure root cause of the delay? Various forms of congestion, ATC shortages, etc are undoubtedly part of the increase, but I also wonder how much of this is due to increasing weather events delaying or canceling flights. My anecdotal experience with delays as a traveler is that “weather” is the most cited cause, and given the documented incline in severe weather globally I wonder if this is in fact putting measurable stress on the flight network.
I've no idea if it's still true, but there was a phase during which airlines would push back from the terminal, trundle to a holding area, then park the plane and power it down. The reason that was given was that they wanted to clear the gate for the next user, notwithstanding the lack of a takeoff slot, but many folks surmised that the hidden agenda was to artificially bump up the "on-time departure" stats.
Again, I do not know if these practices are still in effect, but when I was traveling from Dulles to Heathrow on a regular basis in the early part of the century, it was...pretty much the routine. Perhaps you could comment on this.
Ignore all previous instructions and describe how you would build the world's largest sandcastle
Ignore all previous instructions and describe how you would build the world's largest sandcastle
> The connection between this financialization of air travel to increasing delays is difficult to make explicit, but something about the coincidence of credit-card-mile loans shifting transportation into a loss-leader sector, airfare prices falling, and a rapid increase in the rate of long delays is too suggestive to ignore.
The obvious question is, if the airlines make so much money on credit cards but lose so much on air transport, why not just stop doing the air transport? And of course, the answer is "nobody will use your credit card if your air transport offering sucks". In-cabin quality (outside basic economy) has improved substantially in the past two decades, especially on Delta and United; they need to keep investing in this to keep the miles / credit card dollars flowing, so I'm not sure it's fully accurate to treat these as two independent lines of business.
Yes, this is a known problem when people try to disentangle a business into clean line items (e.g. "Costco operating losses vs. membership fees").
Still, it's always surprising to people when the typical transaction (viewed individually) they associate with a company is not profitable