Mobility and autonomous cars

Around the Bay, and especially in tech, there is a lot of excitement about autonomous cars and what they can do to help people get around. Sadly, the biggest proponents of this new technology are also talking about what it can do to displace transit. I’d like to get into why this is misguided, and talk about what sort of problems we should expect them to solve, and which ones they won’t have much of an effect on.


This is probably the clearest win for autonomous vehicles. As they improve, they’re not going to make avoidable mistakes that human drivers make, such as getting distracted or driving while tired or otherwise not in a great mental state. Autonomous cars aren’t going to drive aggressively, they don’t have blind spots, and they can react to multiple things happening at once. This will help keep drivers from crashing into pedestrians, bicyclists, and other drivers, so we should be happy about this.

Traffic and congestion

People often argue that self-driving cars will be able to squeeze more capacity out of our current road networks. The car’s aforementioned ability to react very quickly and correctly to whatever is going on around it will allow them to pack more tightly on streets and freeways, so people should theoretically be able to get places more quickly.

For freeways, it’s not clear how this is any different from a freeway widening project. We already know these don’t have any long-term effects on congestion because of induced demand, and we should probably expect passengers in autonomous cars to tolerate slower and longer rides than they otherwise would if they were driving because they’re no longer forced to pay attention and drive the whole time. This would make congestion worse. Furthermore, it’s unlikely we’ll have widespread adoption of autonomous cars in the coming decade, and even if we had a sudden, total adoption, the cars will probably need to allow human drivers to override the autonomous driver. This means the vehicles won’t be able to take advantage of tighter spacing and road capacity won’t go up. If we get to the point where the cars never need a driver watching, we’ll have cars on the road without any people in them at all! You might have your car drive you to work, and then go back home to pick up your kids and take them to school. Transit agencies call this empty trip deadheading, and it’s purely a waste of road capacity.

Another argument is that the autonomous nature of cars will allow them to coordinate more effectively among each other. Trusting surrounding vehicles to correctly relay their intent opens us up to a slew of prickly questions about computer security, so I’m not sure we can really evaluate how useful this is, or even if it will happen in the near future. Even if we could, operations like merging onto a freeway involve a lot of actors and necessitate slowing down if the freeway is congested, so there might not be as much room for improvement as we expect.

Street grids will fare a little differently. Since the alleged capacity expansion happens on all streets in the grid at once, we might not see a huge increase in congestion on only major streets as we would with freeways. We should still expect induced demand, and there are plenty of other reasons why driving on street grids will be slow, since we will still have lights and stop signs and pedestrians.

We should definitely not count on autonomous car fleets to take the place of mass transit, as this would entail a huge decrease in the number of people we could move around given the street grid. This is especially important because the total capacity for people to get around acts as a cap on the density and amount of activity the city can support. New York can only operate because the subway moves millions of people per day. LA, on the other hand, has issues with sprawl and road capacity. After decades of trying to build its way out with freeways, it has decided it needs mass transit infrastructure to keep up with its growth and densification. SF’s Subway Master Plan is an admission that getting more people around is ultimately going to follow from investment in transit.


As soon as we get autonomous cars that don’t need a human backup, our demand for parking should become a lot lower, or at the very least, parking won’t need to exist downtown or in other areas where the land could be put to better use. This sounds great, but it could end up backfiring and making congestion and traffic problems worse as a result.

We already know there is a “shortage” of downtown parking, and so if some users switch to autonomous cars that can park themselves further away, it’s unlikely to immediately result in a reduced usage of downtown parking. Plenty of other people are ready to use the parking space that an autonomous car rider would have used otherwise. Even if that weren’t the case, we’d still have to worry about induced demand, and we can model this by looking at what would happen if we added a bunch of free public parking lots. As it stands, the cost or hassle of parking downtown will make transit a more appealing option for some riders, and reducing this burden can get people back into cars, which will cause more downtown gridlock.

We don’t really need to do guesswork here, though—we’ve already seen what happens when there’s a massive expansion of car riders who don’t need parking. In the past, if you wanted to take a car somewhere but didn’t want to park or drive, you would take a taxi. Cities like San Francisco regulated the number of taxis on the road by offering medallions, without which taxis could not operate. The rise of ridesharing companies like Uber and Lyft has effectively lifted the limit on the number of cars used in this manner, and as a result we’ve seen worse gridlock in cities where it is heavily used. As expected, downtown parking is still impossible to find, even when lots of people are riding in someone else’s car.

It’s also worth thinking about what we can do if the demand for parking does drop significantly. Free street parking has already done lasting damage to the built environment in many places. I’m not sure what the best use of the newly freed up space is: road diets might be useful and could free up some more space for buildings (though it would take years for them to be rebuilt to incorporate the space); we can build parklets; some might be replaced with protected bike lanes; some could be transit-only lanes. I’m worried that downtown commuters would want them to be turned into extra lanes of mixed traffic, which would not improve the streetscape at all.

Cost of city transit services

It’s possible that this can reduce operational costs for transit and paratransit systems, but I’m not optimistic about this. Transit operators do more than drive; they help people with disabilities board and alight, and are responsible for resolving situations where e.g. a passenger is harassing other passengers. Even if there were other ways of guaranteeing accessibility or safety, we’d probably still have bus operators. Indeed, trains are now controlled automatically in many metro systems, but we still employ train operators on just about all of them.

As with all new technologies, it’s hard to account for everything ahead of time. But even if these cars do improve traffic and parking, it would just be a band-aid over the deeper issue of decades-long overinvestment in roads and free public parking and subsequent underinvestment in transit and other infrastructure, and putting a band-aid on a problem like this risks entrenching it even further.