On what conditions should New York welcome robotaxis?
In some sunbelt cities, robotaxis are becoming part of everyday life. In Atlanta, Austin, Los Angeles, Phoenix and San Francisco, you can ride a Waymo to your destination with no human behind the wheel. But not yet in the Northeast.
In New York, robotaxis are stalled. In January, Gov. Kathy Hochul proposed legislation to authorize pilot programs for robotaxis. Her plan would have required companies to obtain the approval of local communities before deploying robotaxis, and it excluded New York City entirely. But even this incremental approach proved too controversial. In February, Hochul withdrew her proposal.
At the same time, there is a bill moving through Congress that would preempt much state regulation of robotaxis and other forms of automated driving. A House committee already voted to advance the bill. If the bill becomes law, it could deprive states and cities of their say over many of the rules that govern these new technologies on their streets.
As one of us recently testified to Congress, we believe both in the potential of automated driving and in the imperative of community control. Robotaxis could transform cities for the better — but only if policymakers steer them in the right direction. In a new paper, we argue that robotaxi policy should have three goals: (1) improving safety, (2) reducing externalities and (3) protecting the public.
First, robotaxis must realize their potential to make the roads safer. Robotaxis can avoid many of the errors that lead human drivers to crash. When designed and deployed carefully, they follow speed limits, obey stop signs and traffic lights, and yield the right of way more consistently than human drivers. But they can also make errors that most human drivers wouldn’t make. They have driven around stopped school buses, steered into flooded roads and crashed into a telephone pole.
Credible peer-reviewed work by Waymo, the leading US robotaxi company, suggests that its robotaxis are involved in far fewer injury crashes than humans driving in comparable conditions. Independent safety researchers should thoroughly engage with Waymo’s data, methodologies and conclusions. And it’s too early to say that robotaxis are saving lives: Even Waymo has not driven enough miles, and one catastrophic crash could negate years of safety statistics. But these early results are encouraging, particularly given the tragic status quo on our streets.
State legislators can help regulators and the public learn more about robotaxi safety by requiring that companies report crashes. The National Highway Traffic Safety Administration (NHTSA) already mandates crash reporting at the federal level. But NHTSA lets companies redact important information. New York can fill this gap by requiring companies to report crashes and miles driven not just within its borders, but in any state or country where they operate. And for serious crashes, regulators should require companies to share video, with personally identifiable information obscured. As recent events have illustrated, video has credibility that a narrative crafted by an interested party can’t provide.
Then, state and local regulators should use crash reports to monitor safety. If a crash suggests deeper problems with a company’s technology, regulators should investigate. If they determine a company is creating an unreasonable risk, they should order it off the roads.
Safety is about more than just avoiding crashes. For example, in December, a blackout in San Francisco caused many of the city’s traffic lights to go dark. As a result, robotaxis got stranded on the streets. Waymo explained that its software was designed to check in with a remote assistant before proceeding past a non-functional traffic light, and the blackout caused the remote assistants to be overwhelmed. This incident shows that regulators need to stress-test companies’ emergency planning proactively. How will a company respond in the event of a hurricane, a riot or a cyberattack? Will their fleets be sufficient to help in an evacuation?
Robotaxis are not really “driverless”: Their drivers are the companies that develop and deploy them. State regulators accustomed to licensing human drivers should similarly have the power to license these corporate drivers, to enforce traffic laws against them — including fines — and to require a sufficient corporate presence in the state.
Second, regulation should ensure that robotaxis don’t impose unfair costs on third parties. Here, regulators face hard tradeoffs. All motor vehicles cause pollution and congestion. The deployment of robotaxis could be an opportunity to reduce those externalities because it’s politically easier to regulate a few companies deploying a novel technology than millions of drivers who are accustomed to roads being free. That’s why states have been able to require robotaxi companies to have more insurance than most human drivers would. But at the same time, imposing costs on robotaxis but not personal vehicle owners could distort consumer choices. Robotaxis should be able to fairly compete against other modes, including, eventually, individually owned automated vehicles.
On pollution, the good news is that today’s robotaxis are all battery electric vehicles. The bad news is that, even though electric vehicles have no tailpipes, they still generate pollution from upstream energy use and tire emissions. We would mandate that all automated vehicles use batteries or an alternative fuel to prevent backsliding and then use broad-based environmental regulation to address these other emissions sources.
Congestion can’t be “solved,” but robotaxis could help or hurt. Robotaxi fleets should be utilized more productively than ordinary cars — they shouldn’t spend most of the day depreciating in parking garages. But if robotaxis make car travel more convenient, there will be more car trips. Policymakers should therefore integrate robotaxis into congestion pricing systems, like New York’s successful program, treating them like regular taxis or Ubers. They should also charge all vehicles-for-hire a per-mile fee between rides, so the companies internalize the costs of circling the streets waiting for passengers.
Robotaxis could also create a new kind of externality — surveillance. The sensors they use to perceive objects in their environment passively record their surroundings. And that means every person within range is being recorded too. In some cases, that sensor data could help cities detect potholes or investigate crimes. We have no objection to companies disclosing data when investigators obtain a valid warrant. But we are concerned that robotaxi companies are already sharing video data voluntarily. Companies have strong incentives to do favors for law enforcement and little incentive to protect bystander privacy. Policymakers need to set rules about how robotaxi companies can use privacy-sensitive data, when they can share it, and when they must delete it.
Third, regulators need to protect the public, including those who want to ride and those who do not. The easiest way to keep prices low and service quality high is to encourage competition, both among robotaxi companies and across travel modes. If Waymo raises its fares today, riders can easily switch to a bus, a taxi, or an Uber. But what if fares fall and robotaxis start to outcompete other modes? The transportation market could become highly concentrated. Robotaxis benefit from network effects — the more riders on a network, the less time each vehicle spends deadheading between rides. So larger networks can be more profitable. That’s why the ride-hailing market became an Uber-Lyft duopoly. A similar dynamic could emerge for robotaxis. How can we protect riders if a few robotaxi companies come to dominate our transportation system?
Imagine you had a friend who was moving to New York from Texas. Suppose you were trying to convince her that she wouldn’t need a car to live here. You would explain that the subway runs 24/7, that it covers most of the city, and that the fare is always three dollars. You could say that because the subway is a public service, the MTA isn’t trying to squeeze every last penny out of its riders. But a private company might stand to gain from taking advantage of riders who come to rely on its service. Indeed, there is evidence that Uber’s increasing profitability has been fueled by algorithms that tailor prices to individual riders.
Regulators should ban personalized pricing for fares in robotaxis as well as conventional ride-hails. Every rider who requests the same trip at the same time in the same mode should get the same fare. Robotaxi companies will almost certainly record every trip's origin, destination, time and fare; regulators should have access to the same data for the limited purpose of detecting personalized pricing. If a robotaxi company acquires monopoly power and starts to raise its fares across the board, regulators should impose utility regulation. We expect that the companies won’t love these proposals, yet we think they might encourage robotaxi adoption. People might be much more willing to give up their cars if they know companies can’t take advantage of them.
Robotaxis should be accessible. The industry has long argued that robotaxis will give new freedom to people with disabilities. Regulators should hold companies to their promises. They could require that robotaxi designs be accessible for people who use wheelchairs and other commonly used mobility devices. Or if that’s impractical, they could impose accessibility requirements on each company, create a market similar to emissions trading, and then monitor wait times and fares to ensure that people who need accessible vehicles experience service comparable to other riders.
Our proposals for regulation don’t address all the policy issues that robotaxis raise, including their effect on employment. It’s also understandable that residents of transit-rich cities like New York are more skeptical of robotaxis than residents of car-dependent cities like Phoenix. We, too, are fans of public transit — one of us commutes on the subway (and the other bikes to work). We doubt that robotaxis will make high-throughput transit obsolete. At least on our current streets, a fleet of robotaxis couldn’t move as many people from Harlem to Midtown on Lexington Avenue nearly as effectively as the subway underneath it.
But not every transit line is like the Lexington Avenue subway. Wide swaths of New York City, let alone other parts of the state, are served only by buses that run infrequently. In these areas, the deployment of robotaxis could expand mobility. And to the extent robotaxis do replace low-throughput transit, state and local governments should subsidize rides for people with low incomes.
In the long run, robotaxis could change the transportation system for the better. They could make the roads safer for cyclists and pedestrians. They could expand mobility for people who aren’t well served by transit. They could enable denser development outside of downtowns. And in the long run, they might obviate the need for parking, which will allow us to liberate urban land from cars.
This future won’t arrive if we give the industry a blank check or shut it out. We need to get to work on regulation.
