A flight from New York to Washington is scheduled at roughly eighty minutes. The trip takes four and a half hours. That three-hour gap, not the flight time, is what your travel policy is buying.
Most published rail-versus-air comparisons never close it. They put gate-to-gate against station-to-station, find the plane faster, and stop. Measured door to door, air carries about two and a half hours of overhead that rail does not, and the ranking on short routes flips.
Below is the benchmark we run internally, with the inputs and the assumptions both written down. Copy the arithmetic and run it on the routes your people fly.
What we measured, and what we assumed
The two halves are worth keeping apart, because only one of them is data.
Measured. Rail journey times and departure counts come from Amtrak's published Acela timetable, effective July 1, 2026. We parsed all 176 train columns and counted unique service numbers per day of the week. Amtrak reissues that timetable periodically, so check its effective date before you reuse these times.
Air fares come from the US DOT Consumer Airfare Report, Table 1, fourth quarter of 2025: the average each-way fare paid, all carriers, taxes and fees included.
Assumed, and declared. We model an eighty-minute gate-to-gate block on LaGuardia to Reagan National and LaGuardia to Boston, and ninety-five minutes on Boston to Reagan National. We allow 120 minutes at the airport before departure, which is what the TSA tells domestic passengers to allow, and fifteen minutes from seat to curb with carry-on only.
Airport access is 35 minutes from midtown Manhattan to LaGuardia. Reagan National to downtown Washington is 20 minutes, on the Yellow and Blue lines, which land at a covered walkway into the terminal. Logan to South Station is also 20 minutes, on the Silver Line that Massport runs free out of the airport.
Rail overhead is 35 minutes in total: ten to reach the origin station, a fifteen-minute buffer before departure, ten from the destination station to the meeting. That number is what makes the table below reproducible, so it matters more than it looks.
Every one of those assumptions favors the plane. Closest airport pair, no checked bag, no traffic, no delay.
Three Northeast Corridor routes
| Route | Fastest Acela | Acela departures, Mon to Thu | Door to door, rail | Door to door, air | Air fare, each way |
|---|---|---|---|---|---|
| New York to Washington, 264 mi | 2h 53m | 15 southbound, 17 northbound | 3h 28m | 4h 30m | $202 |
| New York to Boston, 209 mi | 3h 47m | 8 northbound, 9 southbound | 4h 22m | 4h 30m | $203 |
| Boston to Washington, 418 mi | 6h 48m | 7 northbound, 8 southbound | 7h 23m | 4h 30m | $194 |
Rail times and counts: Amtrak Acela timetable effective July 1, 2026, Acela only. Northeast Regional adds materially more departures on all three routes and we have not counted them. Fares: DOT Consumer Airfare Report Table 1, Q4 2025. There is no rail fare column, for the reason given below.
The air column reads 4h 30m three times, and that is not a discovery. It falls out of the assumptions: Boston's longer flight block is canceled almost exactly by Boston's shorter airport access. What it does show is the shape of the thing. Across 209 miles or 418, the flying is a small part of the day and the overhead is most of it.
Where the line falls
One number does the work. Air's fixed overhead runs about 190 minutes out of New York and about 175 out of Boston, where the airport is closer in. Rail's is 35. Call the difference two and a half hours.
So rail beats air door to door when the scheduled rail journey is shorter than the flight block plus that differential. On all three routes here the air door-to-door lands at 4h 30m, which puts the tie point at 3h 55m of train.
New York to Washington clears it by an hour. New York to Boston clears it by eight minutes. Boston to Washington misses by two hours fifty-three, and no reasonable change to the assumptions rescues it.
Eight minutes is noise, so New York to Boston is a tie and the block-time assumption is the entire margin. Move it twenty minutes either way and the route changes hands. On the other two routes the same twenty minutes moves nothing, because the margins are an hour and nearly three. The weakest input decides exactly one of the three, and we would rather name which than claim the model holds everywhere.
The threshold shows up in behavior
Travelers found this line long before anyone wrote it down. A dataset of 132 city pairs, compiled by E. Mitchell at hotrails.net and published in Dr Jean-Paul Rodrigue's work on rail market share against air, puts rail above 70% of the market on most pairs once the journey drops under three hours.
Paris to Lyon is the cleanest natural experiment. When high-speed service cut the journey from four hours to two in the early 1980s, rail went from 40% of the market to 90%. Nothing changed about the plane. The overhead was always there; the train got fast enough to expose it.
On price, and the number we could not get
The air side of the price column is solid. DOT publishes what passengers paid, by city pair, every quarter, which is a better number than any fare you can screenshot.
The rail side is not, and we are not going to fake it. Amtrak publishes ridership by route, reported to the last trip. It does not publish ticket revenue by route in anything we could retrieve, so there is no average Acela fare we can defend setting against the $202.
What we have instead comes from booking these routes rather than from any published figure: Acela business class on New York to Washington usually lands above the DOT air average, Northeast Regional below it, and both move with how far ahead you book. If price decides your route, price it yourself on the dates your team travels.
Run it on your own routes
Nine numbers and one subtraction. On the rail side: the scheduled journey, the trip to the station, the buffer you leave, and the walk out at the far end. On the air side: the flight block, the trip to the airport, the time you make people arrive before departure, seat to curb, and the ride into town at the other end.
Then subtract. Is the rail journey shorter than the flight block plus your overhead differential? If it is, a policy that quietly defaults to flying is costing your travelers hours they could have worked. That is before anyone opens the emissions question, which this model does not measure.
We have not run this on European corridors. If you do, expect it to favor rail harder than it does here, because the airports are further out and the trains are faster. That is a prediction, not a measurement. Airports being further out is also why, on those trips, the useful hotel is usually the one within walking distance of the central station rather than the one by the airport.
If you are rewriting a travel policy, the threshold above is half of it. The other half is what the program costs to run, which we went through in our guide to the real cost and tax treatment of employee travel benefits. The Dyme flight tools price the air half on the routes where the train loses.
Frequently asked questions
Why would anyone take the train instead of flying?＋
On short corridors the train gets there first. New York to Washington is 3h 28m door to door by Acela against 4h 30m by air. The flying is eighty minutes. The other three hours are the ride out to LaGuardia, the two-hour buffer the TSA asks you to leave, fifteen minutes from seat to curb, and getting back into town at the far end.
What is the four-hour rule for rail versus air?＋
There is no four-hour rule, or at least the data does not put the break at four. Across 132 city pairs compiled by E. Mitchell at hotrails.net, rail holds more than 70% of the market on most pairs once the journey drops under three hours. Our own door-to-door model puts the tie point on the Northeast Corridor at 3h 55m of rail time, because that is where the train stops beating the flight block plus roughly two and a half hours of airport overhead.
How do I compare train and plane travel time properly?＋
Measure both door to door, from the same start point to the same end point. Comparing a gate-to-gate flight time against a station-to-station rail time is the error nearly every published comparison makes, and it hides about two and a half hours of airport overhead. Add access time at both ends, the buffer your travelers leave, and the walk from the destination station.
Should a corporate travel policy mandate rail on short routes?＋
A threshold works better than a mandate. Set it at the flight block plus your own overhead differential, publish the arithmetic, and let bookers see which mode wins on the route in front of them. On the Northeast Corridor that puts New York to Washington firmly on rail, leaves New York to Boston a genuine coin toss, and leaves Boston to Washington on the plane.
