We Once Had Self-Driving Cars

(Photo: Elliott Brown)

A frequent response to the dysfunctions of American air travel is technological: namely, self-driving cars (also see this article). In a self-driving car, you can relax, even sleep, while being driven safely to your destination at 60 mph. We once had such a system. It’s called a train network.

Compared to air or car travel, a decent train network is cheaper, more environmentally friendly, and quicker. As an example, I’ll compare two door-to-door, city-center-to-city-center journeys.

The first is by plane, from Boston to New York City:

0:40 Get to Logan airport (airports are typically far from the city center)
1:00 Participate in security theater, etc.
1:10 Sit in loud sardine can
0:30 Get luggage
1:00 Get from Newark airport to destination in New York City
—- ———–
4:20 TOTAL

The journey is 215 miles, for an overall speed of about 50 mph. The current cheapest advance-purchase, nonrefundable round trip costs $150; that’s 35 cents per passenger-mile. For comparison, car travel is about 50 cents per passenger mile (the IRS-set reimbursement rate).

The comparison journey is by train from Paris to Lyon (I need a decent train network for comparison purposes). I made this journey when our family lived for a summer in Lyon, and I learned French by street-fighting methods. The ticket told you where to stand on the platform. Two minutes before scheduled departure, the train arrived. You stepped into the train car, found your reserved seat, and, almost before you had stowed your luggage, the train departed.

0:20 Get to station in downtown Paris
0:05 Arrive early to validate ticket
2:00 Sit on comfortable train with leg room, or stand up without hitting head
0:25 Get to destination within Lyon
—- ——–
2:50 TOTAL

The journey is 289 miles, for an overall speed of about 100 mph. A current advance-purchase, non-refundable round trip costs 70 euros or $100; that’s 17 cents per passenger mile, or one-half of the plane’s cost. The train journey is not only cheaper and quicker, but you get more time to think and read: 2:00 on the train versus 1:10 on the plane.

Forget self-driving cars! If we can print trillions of dollars to create moral hazard by bailing out the gamblers who nose-dived the world economy, why not print money to extend and upgrade the rail network? The U.S. and U.K. rail networks were once twice as extensive as they are today.

Paul Carbone

This is some sort of deliberate obtuseness, yes? A train, even a high-speed one, doesn't drop you off at your specific destination, nor does it operate on one's personal schedule. And then you'll still need transportation to and from the train station itself!

Unlike the route of my car (self-driving or otherwise), I have no control over the rail network. Unlike the car, I can't use the train to pick up groceries or pick up my child from school--and trains are especially poorly equipped to serve unscheduled spontaneous needs. Why on earth would I prefer the inflexibility and inconvenience of a train to a self-driving car responsive to my specific needs on any given day?


Nor does an airplane drop you off at your specific destination. That's the real competition here, not self-driving cars.


While I fully agree with the need for improved train service in North America (I live in S.W. Ontario), automobile travel becomes much more economical when occupancy rates are increased to more than one passenger. For distances that can be covered in less than one day, per passenger cost would be in the 13 cents per mile range.


I have one problem with your comparison. The IRS per-mile reimbursement rate (currently 50.5 cents/mile, IIRC) is much more than actual vehicle expenses, at least for those of us choosing to drive non-gas guzzlers. My own expenses aren't even 10 cents/mile. (Which was great when I got to deduct business travel.)

Alan T

I once took Amtrak from Davis, CA to Richmond, CA. I had an entire train car to myself. The cost, per passenger-mile, of this trip must have been very high, and my ticket certainly wasn't cheap.

Why don't we reduce ticket prices for train travel until all the trains are full? Depending on elasticity, this might require additional government subsidies, but it would reduce the cost per passenger-mile of train travel below that of auto travel (not because tickets would be cheaper but because trains would be full), which would provide a net benefit to society.

In an earlier Freakonomics post, transportation scholar Eric Morris disagress with me, but I don't understand why. Can any reader point out a flaw in my argument?

Alan T

After commenting, it occurred to me that my scheme might increase the total cost of travel to society by increasing the total amount of travel. If we assume that this is not a problem (perhaps because the benefits of travel outweigh their costs), is there anything else wrong with my proposal?

Rhys Daniell

You made your case, but you are being a bit mischievous.

A point to point journey will include transport to the station or airport, and since few stations have large parking facilities they require they require some form of public transport to and from.

A true (automotive) self-driving car does not require a traverse to get to the station, and (assuming widespread use) will be both substantially faster and more economical than current road travel. It will also eliminate delays while you wait for your needs to coincide with the timetable.


Did you discover the break-even milage point where trains (based on your assumptions on wait, travel to the airport/station, security wait, etc.) were overtaken by air travel?


Although the IRS remburses over 50 cents per mile, the true cost of driving (not counting the costs of ownership) is ~11 cents + gas. on an efficient car that can be ~ 20 cents per mile. Ownership is 3-6$/day for a normal car

Kenn Day

The only real reason we do not have an adequate rail system in this country is that there are economic interests who oppose it. They have successfully convinced the majority of Americans that we don't need a rail system. In spite of this, use of the existing Amtrak has skyrocketed over the past few years.

Ben Webster

While I agree with your general point, Boston was a stupid choice. Logan actually is basically in downtown Boston (right across a narrow harbor), and it's incredibly fast to drive there. There's nowhere in the city of Boston, or even its close in suburbs (Cambridge, Somerville, Newton) where it reliably takes 40 minutes to drive to Logan. Public transit can take longer, but thats true of getting to the train station too.


Apples and oranges. Your comparison uses long-haul train networks but the self-driving car is NOT for going between cities. Its for that boring drive to work that chews up 15-60 minutes each way. In a small number of cities there is the density to support rapid transit that is competitive (or better) than the auto, but the rest of us are on their own.

Rhys Daniell

Those who would scoff at self-driving cars should remember that the same was once said of lift attendants - the idea of a self driving machine which could whisk us safely up and down 100 storeys was simply unimaginable.

On the other hand we're spectacularly bad at driving ourselves - world-wide about a million people die each year in traffic accidents and there are many more serious injuries.

And the train zealots should remember that the great majority of people live, and often work, beyond the reach of the train system. That's not due to any failure by the track builders, it's simply uneconomic to service our urban sprawl. The alternative - to force us all to live like battery hens in towers next to the railway - may be environmentally sound but the social cost is unthinkable.


Are you're suggesting building a rail network that goes door to door? That's a different beast altogether.

If rail is good, more rail is better, right?