The policy implications of self-driving cars. By Matthew Yglesias

28 Sept 

www.slowboring.com

11 - 14 minutes

Ten years ago, we saw a crescendo of hype around self-driving cars that then went into abeyance when the problem turned out to be harder to solve than the biggest optimists had hoped. Elon Musk certainly didn’t help. He salted the earth a bit by being incredibly famous and extremely vocal, continually making inaccurate promises and engaging in deceptive marketing around “full self-driving” for Teslas. But along the way, genuine self-driving cars came to be and are now on the streets of San Francisco, Austin, and the suburbs of Phoenix and Los Angeles.

My understanding is that right now, this is mostly a curiosity. Self-driving taxis aren’t really cheaper than normal ones, and some of the markets where they’re available have very low cab demand anyway.

But the technology is not hypothetical. It clearly works, and while there are some technical challenges to rolling it out further, the concept has been proven. It will take some time to fine-tune and get regulatory approval in various places, and at the moment, it seems that when people think about self-driving car policy, it’s primarily these regulatory issues that they’re thinking about.

But that’s thinking much too small.

The one thing we can be pretty sure of with any kind of “computer stuff” is that the cost will come down over time, so whatever the current economics of self-driving cars, they’re only going to get less expensive as long as the technology works. And the most interesting questions to ask aren’t about the safety of self-driving software (obviously there should be safety standards and liability rules), they’re about how we conceive of a world that’s full of really cheap taxis.

I think a lot of people imagine this world of self-driving cars working exactly like today’s transportation system, except that instead of having your hands on the wheel (which some people would presumably enjoy, at least some of the time), you could engage the autopilot and your car would take you to your destination.

This would have various benefits — you could spend extra time doom scrolling on Twitter or reading The Brothers Karamazov, and hopefully the self-driving cars would be safer on net than the human-driven ones.

These would be important benefits, but there’s a bigger picture to consider. Imagine traveling back in time to the 1880s with all the knowledge needed to build modern cars, but you also had to bring modern NIMBYism and hostility to altering the built environment along with you. You could obviously make a big splash in the world and change a lot of people’s lives by building cars and trucks and driving them slowly across uneven terrain and bad roads. In parts of the world where there’s no asphalt and no freeways, people still use mechanized transportation. But the fact is that a world of modern cars and preindustrial roads sucks. You can’t go very fast, everything is extremely uncomfortable. Lots of popular vehicle types don’t really work. And the benefits of modern vehicles to that kind of world would be relatively small compared to the benefits in a world with modern vehicles and modern roads.

Self-driving cars present a similar opportunity.

The big benefit isn’t that you wouldn’t need to drive your own car, it’s that you wouldn’t need to own your own car. Right now, a huge amount of financial value and physical space is taken up by cars that are mostly not in use. That’s because cars are extremely useful, and most people who can afford one want to have a car available at their fingertips. But if the cars don’t need drivers, they can be in more or less constant motion, which means we’d need a vehicle fleet large enough to meet peak demand but not the incredible redundancy of today’s fleet.

The cars also don’t need to be nearly as big on average. If I look out at 14th Street at rush hour it’s full of cars, but the cars are mostly empty. People want to own a car that they can carry passengers and bags for a family trip, but what they’re actually doing is driving alone to work. If you subscribe to a driverless transportation service, then you can ride to work alone in a solo pod, and then take a big family vehicle when you need a family vehicle. Urbanist Twitter likes to mock guys who drive giant trucks to pick up two bags of groceries. But my father-in-law does, in fact, need his giant truck for work. It’s just still also true that this means when he goes to pick up two bags of groceries, he’s driving a giant truck.

Given today’s technology, a fleet consisting of tons of large vehicles that are mostly unused (and mostly empty even when in use) is genuinely the optimal solution. Driverless vehicles should let us move to a much better equilibrium, but that requires not only driverless tech but complementary policy change.

Almost every jurisdiction in America requires new construction to include prescribed amounts of off-street parking.

This is not a good idea under any circumstances. But in most places, it’s not a huge deal in practice. The market demand for parking spaces is genuinely high, and in most of the country the price of land is pretty low so it’s cheap to provide the parking.

Nonetheless, I think cities and states should reconsider this — even in the most car-oriented places on earth there’s no good reason for a regulatory mandate to provide parking — but the rise of driverless technology greatly raises the stakes here. Right now, for example, Kate and I own a car. We could save some money by giving it up, but far and away the biggest cost of car ownership that we incur is the opportunity cost of sheltering our car in our garage. The Logan Circle neighborhood is very expensive, and if we had a little ADU out back to rent out, we could earn a lot of money. But that’s illegal. Even just converting it into a heated space that would be part of our house is illegal. D.C. law requires us and all of our neighbors to maintain off-street parking.

As it happens, we live in a walkable neighborhood with pretty good mass transit, so it’s a place where I think parking reform would lead to a non-trivial fraction of people giving up their cars.

Right now, though, there are tons of people in the United States who currently live in places where land is expensive but the transit options are bad — think the upscale suburbs of every coastal city — and the rise of cheap self-driving taxis could be a game-changer for those kinds of places. The problem is that it’s illegal to make the changes to the built environment that would complement the shift in transportation technology. Even setting residential areas aside for a moment, in pricey areas the natural thing to do as cars become self-driving is to start turning parking lots in commercial districts into cheap apartments. Under current law, though, that’s illegal.

Because places like this are so much more numerous than places like Logan Circle, the economic benefits of parking reform soar in a world of self-driving cars. Or to put it another way, the economic benefits of self-driving cars soar if we adopt parking reform — and adopt it everywhere, not just in America’s relatively rare transit-rich communities.

Speaking of which, the United States has a handful of cities in which mass transit is a genuine pillar of the metro area’s transportation system.

New York is the key case here — it just would not be physically possible for the people who enter Manhattan daily via subway and commuter rail to do so in private cars. It’s a basic geometry problem that exists in attenuated form in a few other American cities. But there’s a much longer list of American cities that have mass transit systems (normally, though not always, just buses) as essentially a social service for the poor and the disabled. Some of these systems operate in cities that are, objectively, pretty big — San Antonio and Vienna have about the same metro area populations — but don’t have an urban built form. But plenty of them are in cities like Bangor, Maine that are downright small.

These systems all tend to be in a dead zone: ridership is extremely low, which means that the cost per rider is very high, which means that improving frequency or expanding service is prohibitively expensive, which means that the quality of service is terrible, which means ridership stays extremely low.

Driverless technology means that low-demand cities could simply provide a cheap (or even free) “public option”-style taxi service that might be non-luxurious or have long wait times compared to private robotaxis while still being dramatically more convenient and comfortable than today’s low-quality bus networks.

Meanwhile, cities that do have intrinsic demand for mass transit based on urban geometry will be able to run driverless buses with much lower operating costs than today’s transit. That will allow for higher frequency (and perhaps expanded service) and also potentially for the use of smaller vehicles.

I’d expect mass transit as we currently know it to be much rarer in a world of driverless cars — the right choice for most cities will be to completely dismantle their existing bad bus operations.

The prior wave of driverless hype saw a lot of concern about the impact of this technology on jobs.

Former SEIU president Andy Stern made the looming automation of truck driver jobs a centerpiece of his argument for a universal basic income program. Today, with the economy enjoying full employment but non-enjoying above-target inflation and high interest rates, I hope it’s easier to see that productivity improving technology is good.

After all, there’s currently a national shortage of school bus drivers. And also a shortage of transit bus drivers.

In a sense, these are obviously not true “shortages,” and if you raise pay enough the vacancies will be filled. But if schools need to massively raise pay to hire bus drivers, that means less money for teacher salaries or for installing air conditioners or improving playgrounds or a dozen other things it would be nice to see schools do. By the same token, we don’t have enough truck drivers largely because this job has a lot of undesirable characteristics. At the same time, compensation for truck drivers flows into inflation and interest rates. And there are tons and tons of vacancies elsewhere in the private sector — albeit a number that is declining due to the Federal Reserve’s deliberate efforts to slow labor demand via higher interest rates.

If we suddenly got cheaper trucking, that would be good (lower inflation). And if a bunch of newly disemployed drivers landed on the labor market, that would also be good because it would allow us to lower interest rates.

Now obviously “you can go get some other job” is not very comforting to a person who loses a job due to layoffs. My mother was a highly skilled analog-era graphic designer who suffered a big hit to the value of her skills thanks to desktop publishing software. The fact that in a technologically dynamic society, a certain number of people are bound to suffer economically from events that are entirely outside their control strikes me as a very good reason to support a generous social safety net. But from the standpoint of society as a whole, it’s good that desktop publishing happened. It’s good that we got digital photography, even if it led to fewer jobs in photo development shops and had a negative impact on employees of Kodak.

The point isn’t to be callous about specific families or even whole communities who suffer from these blows, but that we shouldn’t see technology as a cause of mass unemployment. Mass unemployment was happening 10 years ago because we had bad monetary policy. Today we are suffering because we cannot produce all the goods and services that we would like to have, and if technology lets us get by with fewer workers in some areas, they can be employed in other areas.

The goal is to manage technology-induced transitions in a humane, sensible way — with a welfare state — while welcoming the transitions themselves as our best possible opportunity for sustained economic growth.