There Will Still Be Private Cars in the Future

Robotaxis will complement, not replace, car ownership

As a reminder: Changing Lanes will begin reserving most of its content for paying subscribers in May. Pledge to pay now, at this link, and enjoy Charter Membership rates of $8 per month or $80 per year; that's a 20% discount on regular rates, which will begin on 6 May 2025.

Back in 2017, James Arbib and Tony Seba published a white paper called Rethinking Transportation 2020-2030. That report boldly predicted that within ten years of regulatory approval of autonomous vehicles, which they anticipated happening nationally in the USA around 2021, a new ‘Transport-as-a-Service’ model, principally relying on cheap robotaxis, would deliver 95% of passenger miles in the U.S.

They also projected that robotaxis would be four-to-ten times cheaper per mile than buying a new car, and households would abandon privately-owned vehicles en masse, with the result that the U.S. vehicle fleet would shrink by 80% to just 44 million vehicles by 2030, with only 26 million automated vehicles needed to provide most transportation.

Call this the ‘triumphalist vision’: the idea that as driving automation emerges, it will lead to the frictionless emergence of gigantic robotaxi fleets in major cities, with massive and immediate takeup, and private car ownership will melt like an ice cube next to a roaring flame.

I was thinking about this recently, on my way to a Toronto gaming convention with my son. The con was held at a hotel in downtown Toronto easily accessible by subway, but I chose to drive both days, despite the cost of parking and the sheer aggravation of downtown driving. I did this because the car allowed for secure storage of our things throughout the day: coats, board games we brought that we didn't need at that moment, and acquisitions from trades and vendors. Even in a robotaxi-everywhere world, I still would have made this choice, because robotaxis could not provide this service for me.¹

Arbib and Seba anticipated that by 2025 the triumphalist vision would be well on its way to coming true. It has not, of course.

Nor will it. That doesn’t please me, because the triumphalist vision makes for compelling reading. I’m not immune to that excitement; in fact, I’m particularly susceptible to it, given that I share the belief that driving automation will transform transportation, and eventually society.

But I also understand human nature in a way that triumphalists seem not to. The narrative that we'll all abandon car ownership for robotaxis rests on a series of mistaken assumptions about human behavior. Triumphalists underestimate the average person’s psychological attachment to the freedom of personal vehicles, and fail to consider the practical limitations—like those I faced regarding the gaming convention—that will constrain how quickly and thoroughly robotaxis can replace them.

The robotaxi revolution is coming, and I welcome it. What's holding it back, though, is not the technology, which is impressive and advancing rapidly. There are a host of other factors armed against it. In my book, The End of Driving (coming in August; pre-order now!) my co-authors and I discuss these at length, as well as how to mitigate them.

Today, we’ll consider one factor in particular: the fact that robotaxis cannot serve everyone’s transportation needs. Futurists should instead prepare for a more nuanced future, where robotaxis and personal vehicles coexist for decades to come.

Four Critical Anxieties

Let's begin with the psychology.

There are four types of anxiety that consumers are likely to experience as automated vehicles, whether as robotaxis or as personal vehicles, enter the market. These anxieties will influence both the speed at which consumers embrace automation and their willingness to rely fully on shared fleet services.

Safety Anxiety is the concern that users won't be able to protect themselves from risks when they can't directly control their vehicle. Consider the need to evacuate during a natural disaster, like last year’s Los Angeles wildfires. In such a situation, automated vehicles would struggle with unpredictable road closures, poor visibility from smoke, and the chaos of emergency traffic. An automated driving system (ADS) faced with these conditions would likely proceed with extreme caution or simply fail to operate, precisely when speed and adaptability are most crucial. Worse yet for people who went car-free in favour of relying on robotaxi service; all of these issues would apply, and on top of them, if everyone in their neighbourhood was attempting to flee, it’s unlikely a robotaxi would even be available.

It’s true that most people rarely face such emergencies, though they are becoming more common. Even so, we know very well that North Americans typically shape vehicle purchases around edge cases. Most Ford F-150 owners rarely haul cargo, and many non-EV owners rarely take long road trips, but they all want the option. The same psychology applies to automation: the possibility, however remote, that a steering-wheel-less vehicle might leave us stranded in a crisis is enough to maintain demand for human-controllable vehicles.

A related phenomenon is Control Anxiety, the discomfort with becoming dependent on a machine in such a way that autonomy can’t be regained. It’s distinct from Safety Anxiety in that some people refuse to put themselves in situations where they have to rely on external forces. In its healthy form, this is a fear that they’ll lose important capacities. In its unhealthy form, it’s a manifestation of other fears; you’ll recall that George Will regarded trains as an expression of malevolent collectivism. Rightly or wrongly, this view is deeply held by some, and it’s one that the automotive industry has nursed, at the cost of billions and for more than a century, by championing driving as an expression of personal freedom.

A more practical concern is Privacy Anxiety. The nature of an ADS is that it is constantly monitoring its environment and its location; it can’t operate otherwise. Add to that that robotaxis will, as part of the business model, constantly monitor the car’s interior as well, to ensure that passengers don’t abuse or vandalize the vehicle. Most people will find this surveillance innocuous, or even a safety feature. But some people will see the fact that such cars maintain a record of their trips, and sometimes the behaviour of the occupants, as invasive, or even a threat, and will refuse to participate.

Finally, there’s Access Anxiety, a well-founded concern about the potential limitations of an ADS: whether it might be unable to reach certain destinations due to driving conditions, programming constraints, posted restrictions, location, road conditions, or weather. This parallels the ‘range anxiety’ that has significantly hindered takeup of electric vehicles, and may prove even more persistent. In the much-talked-about SAE Levels, a partially automated vehicle—Level 4—can travel to many places, while a wholly-automated vehicle, Level 5, could travel anywhere. The then-CEO of Waymo, John Krafcik, observed in 2019 that such a vehicle will never exist. Never is too certain for me, but I do think it’s reasonable to believe such a vehicle is decades away, even using the tools we have now; there are just too many edge cases to expect a car to be capable of handling all of them. And until there is a true Level 5, the people who want assurance that they can reach anywhere they want to go will want a car they can operate themselves.

Underlying all these anxieties is that old devil, familiar to us from behavioural economics: loss aversion. Simply put, people prefer avoiding losses over achieving equivalent gains. For consumers, it will loom large in decisions about private vehicle ownership. Rather than bloodlessly weighing the benefits of owning an automated vehicle versus its costs, they will pay close attention to what they might lose: safety, control, privacy, travel options, and even identity. Consumers considering giving up car ownership in favour of just calling a Waymo when they need to travel by car will feel the bite of loss aversion all the more.

Some of these anxieties may diminish naturally over time. History shows that we generally grow to trust technologies with familiarity and repeated exposure, and those who grow up with a technology take its benefits and features for granted. I’ve never ridden in an elevator that had an operator, for instance, and I’ve never felt the lack. But familiarity and repeat exposure, much less generations that grow up taking it for granted, will take time to develop.

Consider this scenario: It's 2032, and you're in an automobile showroom, comparing two options for your next car. The first is conditionally automated, and can drive itself, not only on the highway, but between any two addresses in the city… but it does have a steering wheel, and it would expect you to take over in situations it can’t handle.

The second is completely automated, so much so that it lacks any controls. It will never need you to take over because it’s that good. While driving in your town and on the highway, it will always handle things, leaving you to focus your attention on other things. Unfortunately, it can’t operate in extreme weather, so you won’t be able to use it during a severe storm; and if you’re on the road at that time, the car will pull over and you’ll have to wait till the skies clear again. Further, it won’t travel on the unpaved roads to your sister’s cottage. And on the way there, it won’t pass through some of the small towns on the route, because they banned automated operation after a fatal accident.

For many people, the choice of which car to buy is clear. It’s not the one triumphalists predict.

Practical Barriers

People will have complicated feelings about driving automation, but those aren’t the only problem the triumphalists have. Psychology isn’t the only barrier to relying solely on robotaxis. Some people have fundamental demands (or, sigh, ‘use cases’) of personal transport that robotaxis will struggle to accommodate.

Take the situation of families with young children. Young children need to be restrained in car seats that are customized for their age and weight; it’s not just a good idea, in most jurisdictions it’s the law. Take these regulations, and add parents, who will want access, on demand and right away, to a vehicle with suitable seats that are safely configured, familiar, and clean. Most such families, given the opportunity, will prefer ownership of a family vehicle. It’s hard to imagine any robotaxi firm, no matter how many vehicles they have on the streets, being able to meet the family’s expectation in a way that is equally good and cheap as private car ownership… much less better.²

People with health and accessibility challenges are in a similar situation. Many such people require special fittings, sizes, egress, loading areas, and occasionally human assistants to use a vehicle. It’s easier to accommodate these needs in a private vehicle.

Pet owners face similar challenges; pets, especially dogs, can leave behind odors, dander, dirt, and waste, and can damage interiors. Consequently, robotaxi operators may decline to permit pets in their vehicles. If this is so, pet owners will continue to own their own.

Workers are another segment. Gardeners, plumbers, electricians, construction workers, and dozens of other service providers carry tools throughout the day. Sales representatives often transport samples, equipment, and signage from site to site. And any private individual who runs multiple errands, such as picking up groceries and dry cleaning, needs storage space between stops, just as I did when my son and I went to the gaming convention. For all these individuals, a personal vehicle with dedicated storage capacity remains essential.

The cumulative impact of these barriers suggests that many households will maintain at least one private vehicle even in a future where robotaxis are commonplace. The triumph of the robotaxi will be partial, not total.

Users Care About Their Needs, Not the System’s

Imagine that every second Friday, a divorced father drives from the mid-town area of a large metropolitan city to the periphery, to collect his daughter for a weekend visit. The journey involves multiple stops: first to her mother's home where the daughter boards with an overnight bag, schoolbooks, guitar, and other items; then to pick up her older sister at another location; next, to a restaurant; back to drop off the sister; grocery shopping; and finally to the father's home.³ This complex choreography of people and cargo illustrates why many regular journeys are better accommodated by a private vehicle. Breaking it into multiple robotaxi trips would compromise nearly every element of what makes private vehicles attractive.

The inconvenient truth that these two stories illustrate is that there is a strong disconnect between what planners want out of the transport system and what users want.

Planners want a world where the vehicles that constitute the transport system are Automated, Connected, Electric, and Shared (ACES). ACES describes a subway system today, and the robotaxi network of our dreams. Such a system is maximally efficient, and as such is best for all travellers, in aggregate.

But consumers don’t regard the system as a whole. They regard their own case. And the revealed preference for such consumers, overwhelmingly, is for vehicles that are Comfortable, Affordable, Fast, and Instant, i.e., instantly available (CAFI).

Right now, in almost all places, private ownership is the only way to achieve the CAFI preference. None of transit, carpooling, and ride-hail are able to do so. I expect that, absent significant policy intervention to change the outcome, robotaxis will join the list.

Robotaxis will thrive in cities, and serve riders with straightforward needs. They will erode transit ridership, certainly, but they will not, in the foreseeable future, eliminate the need for private cars.⁴ Households with children, tools, pets, or complex routines will continue to own personal vehicles. These vehicles will almost certainly be electric; they will be somewhat connected and somewhat automated, but they won’t be shared.

The triumphalist vision makes for compelling storytelling. But the future of transportation will be determined, not by the technology, but by messy human needs and desires. Robotaxis will certainly transform urban mobility, but they'll do so as one important component in a diverse ecosystem, rather than as the overwhelming force that triumphalists predict. Revolution is fun to read about, but place your bets on evolution.

1. It’s worth noting that even if public transit was ten times better than it is today, so more frequent, faster, and cheaper, I still would have chosen as I did.

2. But see here for a strong counter-example. The family in that story didn’t own a car, and one reason they preferred a Waymo to an Uber to bring their newborn home was they knew the Waymo would wait patiently, as long as it took, for the inexperienced parents to properly secure the car seat. It’s true that robotaxis will be patient in this way, but I nonetheless believe most families in this position will prefer their own car with a seat permanently installed.

3. This is a personal case from The End of Driving, drawn from the lived experience of one of my co-authors.

4. To avoid such an outcome, transit agencies will have to embrace robotaxis. This will be hard, but will offer great benefits, just as their limited automation of rail service has done. The End of Driving has a good deal to say about this; we will explore the matter later this year.