←GRAB→
CityLab
Perspective

Why Robotaxis Can Make Cities Safer

An urban policy analyst at the Manhattan Institute argues in favor of expanding autonomous ride-hailing in San Francisco and beyond.

By Jordan McGillis
September 20, 2023

In August, the California Public Utilities Commission voted to allow Cruise and Waymo — GM and Alphabet subsidiaries, respectively — to operate autonomous vehicles, or AVs, in San Francisco around the clock. The decision has drawn fire from multiple directions, including from urbanists who decry any policy they deem favorable to automobiles. 

One such group, a band of mischief-makers who call themselves Safe Street Rebel, racked up clout on TikTok ahead of the regulatory change by plopping orange traffic cones onto the hoods of AVs trundling about San Francisco, disabling them on the spot. Following the CPUC vote and an uptick in AV-related traffic incidents, San Francisco City Attorney David Chiu joined Safe Street Rebel in spirit, filing a motion to redo the regulatory decision and claiming that the city “will suffer serious harm” as AVs become more prevalent."

But if healthy, functional cities are the goal, opposition to the autonomous ride-hailing service that Cruise and Waymo offer is wrongheaded, despite the recent hiccups. As a policy analyst at the Manhattan Institute, I work at the nexus of transportation and urbanism. My latest report, “Autonomous Now: Why We Need Self-Driving Technology and How We Can Get It Faster,” outlines the specific benefits of autonomous ride-hailing and explains how governments can pave the way for the new technology without unduly tilting the transportation balance towards cars. What these activists and municipal officials are missing is that their core concerns — safe streets, urban quality of life and affordable transportation — are problems that AVs can help address. 

  

The crux of the urbanist case against autonomous vehicles is that they entrench car culture — a set of norms that Safe Street Rebel and its fellow travelers are in many ways correct to lament. In the past three years, America’s car-centric transportation ecosystem has left more than 120,000 people dead on our roadways and has sent as many as 10 million more to emergency rooms. These grim statistics are not strictly a phenomenon of rural highways. America’s five largest cities — New York, Los Angeles, Chicago, Houston and Phoenix — saw more than 1,200 deaths from vehicle collisions in 2022. More than 100 people have been killed on the streets of San Francisco since 2020. Reform is certainly in order, but, contra Safe Street Rebel, autonomous ride-hailing is a solution worthy of inclusion. 

The case for autonomous ride-hailing (henceforth “ARH”) rests on the foundation of the empirical benefits provided by ride-hailing of the human-driver variety. Setting AVs aside, the emergence of on-demand mobility companies like Uber and Lyft has been proven to significantly reduce roadway fatalities and car crash ER visits in cities where its impacts have been evaluated. A 2021 National Bureau of Economic Research paper concluded that the presence of just one ride-hail company, Uber, saved nearly 500 American lives in 2019. The leading causal factor is that ride-hailing cuts down on drunk driving by providing a low-friction alternative to getting behind the wheel under the influence of alcohol. Even with the ride-hailing option widely available, however, drunk driving still kills more than 10,000 Americans annually. 

Autonomous vehicles will multiply the benefits of ride-hailing in two ways. First, within well-mapped and defined parameters like San Francisco’s streets, autonomous driving systems are already showing themselves to be better at averting collisions, particularly those that cause injury, than their human counterparts. Amid skepticism towards ARH safety claims, technology journalist Timothy B. Lee recently analyzed California’s entire database of AV-related incidents. By his count, Cruise and Waymo have reported a combined 102 crashes involving their vehicles over roughly 6 million driverless miles in San Francisco through August 25, 2023. “These were overwhelmingly low-speed collisions that did not pose a serious safety risk,” Lee writes. “A large majority appeared to be the fault of the other driver.”  

Over 2 million driverless miles in San Francisco, Waymo vehicles have been involved in just four major crashes — half as many as would be expected from a human driver at the national average. These figures are in line with a study Cruise produced in conjunction with the University of Michigan and Virginia Tech that concluded its cars are 50% less likely to be involved in collisions and 70% less likely to be involved in injury-causing crashes compared with human ride-hail drivers in urban areas.

Robotaxis Outperform Humans in San Francisco, Cruise Data Suggests

Source: Cruise, University of Michigan Transportation Research Institute, Virginia Tech Transportation Institute
Data comes from preliminary study results published by Cruise. Collisions with risk of injury are defined as vehicle collisions "requiring application of medical examination and treatment."

While AVs will never be perfect, the proper point of comparison is the human-driver status quo, not infallibility.  

Second, autonomous ride-hails have the potential to be much more affordable than current Uber and Lyft fares. The ARH business model differs markedly from its forerunners: The companies own the cars, maintain them, and, of course, pay for the autonomous tech. While this asset-heavy approach comes at a cost, the savings that will be generated by eliminating driver pay more than make up for it. A 2022 McKinsey analysis suggests that, by 2030, ARH rides could be less than half the price of current human-driver rides.

Such savings stand to pay off in several ways. By offering lower prices, ARH will entice more risky drivers out from behind the wheel and into safe transportation. And lowering the entry costs to ride-hailing would have equity benefits that go beyond ferrying Friday night revelers: Non-driving adults like teenagers, elders, people with disabilities and others who now find themselves constrained by the car-ownership norm and underserved by transit will have their access to community life enhanced — a crucial benefit amid our national struggle with social isolation. 

Now let's address some of the other critiques. Autonomous vehicles’ urbanist detractors tend to fear that AV proliferation will take momentum away from reallocating street space toward bikes, pedestrians and less-polluting car alternatives. But the real effect may be the exact opposite. As shared self-driving vehicles chip away at the need for a personal car, more people will begin to think of themselves as non-drivers. Once that shift begins, the condition that anti-car advocates have dubbed “car brain” will ease, and with it barriers to car-light urban design. Instead of relying on extensive signage and traffic-calming measures to mitigate the risks posed by error-prone human drivers, we can begin to build with more confidence that cars present a limited threat to people outside of them. What that looks like, exactly, is yet to be determined. But autonomous vehicles can facilitate a streetscape that is more harmonious with people. San Francisco is now on the frontier of what that looks like. 

Another urbanist concern is that ARH will erode public transit utilization. But to the extent that it takes riders away from transit, it will be a reflection of people’s needs being better served by a new model. If ARH gets people from Point A to Point B at prices they find more appealing than transit’s — both in terms of money and time — that should be taken as a transportation success. As urban economics doyen Alain Bertaud has shown, cars are increasingly useful as commercial activity becomes less geographically concentrated within metropolitan areas. Fortuitously, self-driving is arriving just as transit’s viability in many cities has lost steam.

Autonomous vehicles can facilitate a streetscape that is more harmonious with people.

Moreover, the technology girding AVs opens up exciting transit opportunities that do not rely on costly fixed infrastructure like rail. Imagine, for instance, if Hong Kong’s minibus system was adopted in San Francisco or other US cities, but with autonomous technology at the helm. The cost-per-rider possibilities are tantalizing. And indeed, Cruise is working on a shuttle vehicle reminiscent of a minibus that it calls the Origin .  

Rather than warding away autonomous vehicles, governments should establish policies that let AVs work to improve safety and mobility. In my report, I suggest explicitly allocating more curb space for ride-hail pick-ups and drop-offs of both the autonomous and human-driver varieties. Another idea is to give ARH access to some special areas that now exclude private cars but allow transit vehicles, like San Francisco’s Market Street. A third recommendation is to package AV adoption with policies such as congestion pricing, which will cut down on the possibility of driverless “zombie cars” ceaselessly trawling for riders by charging access fees to busy streets. State governments, for their part, should all designate an office that serves as an AV policy hub, as Texas does. The federal government, meanwhile, should focus its resources on accumulating data on AV safety performance and should use its vehicle recall powers where necessary.  

Cities need safer streets and better mobility options. ARH opponents are correct on that, but little else. Far from meriting rejection, autonomous vehicles are a smart solution to pressing urban problems, and the California Public Utilities Commission deserves plaudits for granting them a green light.  

Jordan McGillis is a Paulson Policy Analyst at the Manhattan Institute. He posts on X at @jordanmcgillis.

CityLab
Perspective

Why Robotaxis Have No Place in the City

Two San Francisco activists explain why they're trying to halt the city's autonomous ride-hailing vehicles in their tracks.

By Aditya Bhumbla and Mingwei Samuel
September 20, 2023

Tech fanatics cheered last month when the California Public Utilities Commission voted to allow two self-driving taxi services to massively expand operations in San Francisco. With robo-drivers at the helm, they hope, any destination will soon be a cheap, clean and safe ride away, as autonomous vehicle companies Waymo and Cruise promise.

But we have bad news: That idyllic future is not coming. Cars are fundamentally incompatible with cities, and self-driving cars cannot change that. Instead, they introduce even more problems, which we’ve seen firsthand in San Francisco.

We are members of Safe Street Rebel, a grassroots activist group advocating for safe and accessible streets for people, not for more cars. You might’ve heard about our protests against AVs, where people place traffic cones on their hoods, which obstructs their sensors and immobilizes them. For safety, we oppose coning AVs that are carrying passengers or that are in crosswalks or on bus routes. While these actions do carry some risk, our ground rules serve to minimize that risk while we fight back against what we see as the much greater threat that AVs represent: entrenching a world already dominated by cars.

Robotaxis have caused trouble in San Francisco. Footage courtesy of Safe Street Rebel.

In our real-world experience, AVs fail in many of the same ways human-driven cars do: They fail to yield to pedestrians, block up intersections, and park in bus stops, crosswalks, and bike lanes, which forces vulnerable people out into traffic. But they are also failing in new ways. They frequently block traffic by stalling in place, often near road work, in intersections, or next to other stopped cars. There have been countless such incidents, which shows how truly not autonomous current self-driving vehicles are, since they need teams of remote human operators to get unstuck. 

Our coning protests highlight and demonstrate these problems. But most disabled AVs get stuck on their own, in far more dangerous or inconvenient places. For example, AVs have become a huge problem for first responders. The San Francisco Fire Department recorded 55 incidents of interference with their vehicles, and many more are inevitably coming – which is part of the reason the San Francisco city attorney has filed for a rehearing of the CPUC’s decision. Blocked crosswalks and delayed fire trucks may seem insignificant compared to the graphic violence of car crashes, but these incidents pose a serious safety risk. Should an AV cause a death by delaying a fire truck or ambulance in an emergency scene, or cause one pedestrian to be injured while circumnavigating a crosswalk or bus stop blocked by a driverless car, the whole AV safety pitch comes crashing down. 

AVs fail in many of the same ways human-driven cars do. But they are also failing in new ways.

AVs also drive erratically. Last year the National Highway Traffic Safety Administration launched an investigation into Cruise vehicles’ tendency to brake hard and suddenly, in a few cases causing rear-end collisions. We’ve also seen robotaxis struggle with completely normal traffic situations and fail to adapt to changing road conditions, in one case driving into wet concrete.

Humans are certainly terrible, unsafe drivers. But claims of AV safety are simply not backed up by public data, which is itself sparse and disputed. Cruise, which is owned by GM, has said that compared to humans, its AVs get into 73% fewer crashes with a “meaningful risk of injury.” But San Francisco city officials have calculated an injury crash rate for robotaxis 6.3 times the national (human) average. Cruise’s vehicles only just reached 4 million driverless miles in August, with Waymo, which is owned by Google parent Alphabet, trailing behind. Human drivers kill one person approximately every 80 million miles or so in San Francisco, so AVs still have more than ten times farther to travel to reach that mark. After voting in favor of robotaxis last month, CPUC commissioner and former Cruise lawyer John Reynolds stated “we don’t yet have the data to judge AVs against the standard that humans are setting.” It’s true — which is why we should be wary of these companies’ claims.

Even ignoring these significant shortcomings, we still have a geometric problem: Cars are an inherently space-inefficient way to move people around cities. Dedicated transit infrastructure, bike lanes, or even just sidewalks can move more than ten times as many people per lane.

Ultimately, cars — autonomous or not — will always be wasteful, whether powered by gasoline or electricity. The mining of lithium and cobalt for batteries destroys fragile ecosystems on Indigenous lands and has resulted in worker deaths and violations of human rights. Compared to gas-burning cars, electric vehicles have lower tailpipe CO2 emissions, but they still contribute to particulate pollution from tire and brake wear, which is one of the largest sources of microplastics in waterways and produces more particulate pollution, by mass, than tailpipes

It’s naive to think capitalism will save us. While Cruise and Waymo promise to expand mobility by delivering cheaper rides than human-driven taxi services, their initial fares are no cheaper than Uber and Lyft. AV companies claim their product will lower car ownership, but so did those ride-hailing services, and in their aftermath, car ownership is up. Ride-hailing snatched passengers from trains and buses instead, and robotaxis do nothing to change that. In fact, politicians are already using AVs to justify undermining public transit. And police departments are already using footage from these “surveillance cameras on wheels.” Is this the world you want to live in?

Many AV proponents, including some tech executives and venture capitalists, are quick to paint people resistant to robot cars as Luddites, afraid of technology. While we do share the labor concerns of the original Luddites, the truth is many AV opponents — including us, the authors — are tech workers ourselves. From experience, we know that “move fast and break things” might work fine when creating pure software, where changes and reversions can be made quickly. But where software interacts with the real world, that approach is naive at best and harmful at worst. We should be measured about the technology we add to our cities, not cheerleading anything novel for novelty’s sake.

We don’t need techno-futuristic fantasies to solve our urban transportation challenges. We already know what we need to do: eliminate dependence on cars, period. Our roads can be far more efficient if we dedicate more space to buses, bikes, and pedestrians rather than two-ton metal boxes. We can use materials more responsibly; the lithium ion cells in one EV battery can power 300 e-bikes. Cities designed for people in other parts of the world already have far safer roads than counterparts in the US, no AVs needed. Places like Paris, Jakarta, and Bogota are moving past cars to build healthier, friendlier, more sustainable urban spaces. The US needs to learn from these real successes, instead of using technology to double- and triple-down on car dominance. We won’t end car-dominance by adding more cars.

And what if, when the data comes in, AVs are indeed safer than human drivers? In many ways, that question is missing the point. By framing the conversation exclusively around vehicle safety, we ignore the chance to build a better future through bikes, buses and trains. Companies like Waymo and Cruise are trying to solve car problems by removing the human instead of removing the car. That’s why we’re fighting back.

Aditya Bhumbla and Mingwei Samuel are organizers with @SafeStreetRebel.