In the abstract, self-driving cars should make traffic in cities better. In practice, I think self-driving cars will make traffic congestion far worse, requiring a fairly substantial redesign of the road systems in American cities.
The short-term congestion and long-term time, expense and disruption of the required redesign of city roads present yet another serious challenge for any company hoping to profit from a fully autonomous vehicle future. It will markedly slow down the transition, and give cities a major lever to extract rents from any company hoping to operate unmanned vehicles on their streets.
In a sense, the conversion of American cities to self-driving tech is likely going to feel a lot like a small home renovation gone bad. Once you open up the walls for what seems like a minor change, you find out the whole foundation needs to be replaced, and the project—while still important—isn’t quite as profitable as it may seem at the outset.
The Practical Problem Today
Most people believe that that ride-sharing services are an early proxy for how fully autonomous transportation will work. So, to understand the future problem today, all you need to do is try driving from the Mission to the Marina down Divisadero in San Francisco. The street is a major artery with two lanes in each direction. It used to be that the left lane was the slower lane because you had to wait for people to make a difficult turn across traffic, as is the case with most similar roads.
In the last 18 months, however, the right lane has become completely blocked with ride-sharing cars waiting for passengers along the street. The left lane in each direction is, amazingly, while still slow, now the faster of the two.
The implication is that what used to be a somewhat traffic-laden main thoroughfare is now virtually impassable at the wrong time of the day.
Of course, this isn’t a one-street problem. If all of a sudden you have lanes blocked everywhere with waiting cars that didn’t used to exist, the whole system slows to a crawl. And this is well in advance of the full brunt of what a cheap self-driving version of these ride-sharing services would look like.
The cheaper prices likely to be offered by self-driving taxis mean far more passengers—and vehicles—on the road. If self-driving cars drive further reduction in the cost of driving, people who now take public transit will likely switch to the roads. More passengers doesn’t just mean more pressure on the system. It means far more cars waiting for people on all streets.
One-way streets become impassable if there is just one person being picked up. Multi-lane thoroughfares will effectively lose the rightmost lane to waiting. So, parking will have to be converted into waiting space for self-driving vehicles. New holding areas will have to be designed. And many streets will likely need to be simply off-limits to self-driving vehicles.
In short, hello jackhammers, road closures, and lots of crazy traffic pattern changes before self-driving cars help rather than hurt the flow of traffic in cities. It will take years and amazing amounts of political pain.
Who Pays for the Conversion to Self-Driving-Friendly Streets?
In a world where ride-sharing vehicles and self-driving cars piggyback nicely on existing infrastructure that cities have already paid for, it is reasonably hard for the cities to extract fees from the car services.
However, in a world where cities need a major retrofit in order to support new traffic patterns, there is going to be a real question about who pays.
A few generations ago, the auto manufacturers of the day struck an incredible deal with the U.S. taxpayers. The public more or less footed the entire bill for the rollout of the highway system, which made cars useful and valuable.
I don’t think it will happen the same way this time. Very wealthy companies are getting into the self-driving car game, and most municipalities are relatively poor. It is likely, therefore, that cities will look to the companies offering self-driving systems to help foot the bill for the congestion and retrofit work they cause.
It could take the form of cities charging a very high annual fee for each self-driving vehicle on the road or automatic congestion fees. There could be rules requiring self-driving cars to use specific loading zones, created by auctioning off existing parking spaces. It could also be more direct transfers of cash or equity.
Of course, this is all before the local governments realize that they hold a valuable chit in a big new market and should try to extract value from the companies even above and beyond the cost of conversion.
But regardless of the details, if directionally things play out this way, it could lead to some very different patterns for who captures the value from the new technology. It might be that being in the self-driving vehicle business isn’t really all that good a business after all the rents and taxes are paid.
Self-Driving Is a Long Way Off…
In 2013 I wrote a post with my guess at the timeline for the rollout of self-driving cars. I made the call that it would be five to 10 years before cars drove autonomously with human backup, and 25 years before fully autonomous cars without human drivers were on the market.
In the last 18 months a lot of people have asked me if I regret guessing it would take that long. I might have been slightly too pessimistic about the fully autonomous timeline for some limited cases on highways and in some regions. But I have actually become more pessimistic about the midterm viability of fully autonomous cars en masse in cities.
It is possible that some new cities will be built differently and fully autonomous tech will roll out faster. But my money says that as you go through the layers and layers of what it means to get self-driving cars working in cities that were laid out in some cases pre-automobile, don’t hold your breath.