Going Driverless, or Not

A heated debate over the significance of Google's so-called driverless car has been raging over the past several weeks. On one side of the aisle are those hailing it as a "revolutionary" technology that will dramatically alter personal mobility to the point of  eliminating private car ownership. On the other side are those who reject the premise that the technology represents a groundbreaking shift, instead characterizing it as merely a "slightly different variation" on current transportation modes that is "so incremental that it epitomizes our national short-sightedness, and failure of imagination, when it comes to improving mobility in America."

It's difficult to imagine two more divergent positions on the significance of a new technology. Although I'm wary of attempting to forecast the future, knowing how likely it is that any predictions are likely to appear foolish or worse some years down the line, there are enough parallels, current and historic, that I think some general observations can be made here without wandering too far off into pointless speculation.

The human-driven motorized car, itself a recognizable variation on the 5,000 year-old horsecart, must have seemed like a rather mundane idea in the late 19th century — almost a throwback to horse-and-carriage travel that the railroads had put out of business — compared to steam railways, attempts at early aircraft and even the bicycle, which represented the first instance in history of a fast yet human-powered wheeled vehicle. The name for the early cars reflected this mindset, which some have used to argue that driverless cars are also being underestimated, as in this blog post:

"One reason I will eventually move away from my chosen name for the technology — robocar — along with the other popular names like “self-driving car” is that this future vehicle is not a car, not as we know it today. It is no more a “driverless car” than a modern automobile is a horseless carriage. 100 years ago, the only way they could think of the car was not notice there was no horse. Today, all many people notice is that no human is driving. This is the thing that comes after the car."

Could it be that these early observers were right, though? The very early car was slower than the steam trains, and its primary breakthrough was economic: by moving the horse out of the picture, and substituting the combustion of an oil, it became possible to maintain a personal carriage without the the constant care and expense needed for keeping horses. The earliest car ads emphasized the savings in cost, care and anxiety from not having to keep a horse rather than advantages in speed. It took several more decades until the velocities enabled by combustion power and paved highways could be fully realized.

By contrast, the driverless car offers no such economic advantage to the individual driver, since he is already donating his own labor to operate the vehicle. In implicit recognition of this fact, the claim is made that the most significant consequence of this invention will essentially be to reduce the cost of taxis to the point that renting a car on a trip-by-trip basis actually becomes cheaper, and no less convenient, than owning one. In other words, a driverless car network, for all the technology it requires, is really a simple labor saving device, which like the very early car, allows an existing function to be performed more cheaply but otherwise not much better: certainly driverless cars will not enjoy the kind of speed advantages over human-driven cars as the autos of the 1920s gained over horse-drawn carriages. In this sense, it is functionally equivalent to a massively subsidized (or, perhaps, completely unregulated) human-driven taxi service, which, in theory, could be funded for no more than the amounts currently spent on private car ownership, and certainly with less technological difficulty.*

As it happens, non-subsidized transit systems of this sort already exist, and have existed for decades, in many cities of the developing world where labor is cheap, car ownership is low and public transit options are limited.

Auto rickshaw in Bangalore and Xe Nom drivers in Hanoi. Wikipedia and Flickr/Gavin White

These transit systems, based first on human-powered rickshaws and later auto rickshaws, motorbikes and pedicabs, provide both point-to-point and last-mile transport, essentially substituting for private vehicle ownership.  However, the tendency over time, as a country grows wealthier, is not for auto rickshaws to become ubiquitous, but for public transit and private vehicles (motorbikes or cars) ownership to supplant them. This is certainly due in part to the rising cost of labor, but must also be due to the inherent comfort and convenience of owning one's own vehicle in low-density areas and of the geometric efficiencies of transit in a dense city (using Jarrett Walker's terminology). The dense city of taxi-based transport tends to be a traffic nightmare. The low-density city, on the other hand, generally uses taxis in a limited supportive role.

(An exception might be the dense but mid-sized city of the developing world, such as Jaen, in Peru, a country that is notorious for the use of largely unregulated, and therefore quite cheap, taxi and bus systems to supplement inadequate public mass transit systems, such as in the far larger capital of Lima):

Motorbike-taxis on the streets of Jaen, Peru. Exploration of the city on Streetview shows a
largely taxi-based transportation system, supplemented with private motorbikes. There are
only a handful of automobiles visible here and there, mostly utilitarian in nature. A far cry
from the SUVs of American streets, these vehicles actually seem designed around the
size and weight of human beings, and create a steady but by no means congested flow of traffic.

For driverless cars to reverse this durable observed trend, and actually encourage people to dispose of their cars, we'd need to believe that the cost savings of driverless taxis could outweigh the inevitably increased inconvenience of not having personal ownership (including ownership of a driverless car) for most or all people. The problem with this scenario, though, seems to be twofold, as stated before:

• In dense urban areas, very cheap and convenient taxi service may overwhelm highways and city streets, as it does in Hanoi or Mumbai or many other cities of the developing world, negating that same convenience and worsening the quality of urban life (miles driven are expected to increase with a "robotaxi" system, according to one recent study, and this may underestimate the number of transit riders, cyclists and pedestrians who switch to driverless taxis). If the number of taxis is limited, as in most cities in developed countries today, then this will partially negate the cost advantage of lacking a human driver and will certainly hinder convenience, illustrating again that this is as much an issue of restrictive licensing and geometry as it is technology. Arguments on behalf of robotaxis appear to assume that their numbers will not be limited by law.
• In suburban and rural areas, sufficiently frequent robotaxi service may be difficult to provide, yet the cost of storing one's own personal vehicle will continue to be minimal or nonexistent while providing total convenience. Additionally, suburban errands often require multiple stops spread out over a large area due to car-based urban design, which will either require tiresome and inconvenient re-hiring of cars for each leg and practical difficulties with transferring purchased goods, or else cars will need to 1) park at each destination or 2) cruise around aimlessly while waiting, either of which would lessen certain of the the advantages over individually owned vehicles. 

There is one other issue which I raise by way of a insightful quote from Neil Salmond's article on robotaxis: "Once you own a car - and so mentally discount the cost of insuring, maintaining, fuelling the car - then every trip looks free." Although I agree with this, it also suggests a very difficult path toward weaning people off private car ownership. A privately-owned car is immensely versatile. It can handle short trips, medium-length trips and cross-country treks. It is always there and ready to go at a moment's notice, at times of high and low demand alike, with no questions asked. For many it is a personalized space as much as one's own bedroom or office, and doubles as a mobile storage unit. The opportunity cost of giving up all this "free" travel, and its associated comforts and conveniences, may therefore be much higher than might be suggested by a pure dollar-and-cents comparison.  Further, once the car is owned, whether it be human-driven or not, the incentive to use a taxi system at all is much reduced for the very reason Neil describes. No matter how reasonable a robotaxi trip might be, it cannot beat the perception of "free."

Whether or not robotaxis can succeed at large scale in American cities, there are a few areas where a privately-owned driverless car could provide an unmitigated social and economic good. For those unable to drive and without access to reliable transportation, such as young people or very elderly living in car-dependent areas, owning such a car could be a lifeline to mobility and independence. That would have the potential to remedy one of the greatest inequities of a car-based transportation system (though certainly not the only one). It's also tempting to imagine, as Neil suggests, driverless cars ferrying commuters to rail stations, and thereby opening commuter parking lots for redevelopment, and it's certainly possible, or even likely, that driverless cars could become the default ownership option. A cheaper, but non-revolutionary, taxi system might not be such a bad thing, especially for households that need access to a car but only rarely. And what of the implications of driverless buses, and driverless car sharing? As an incremental step that expands transportation options while lowering costs, it has promise.

The vision of a fleet of driverless taxis completely supplanting car ownership, however, seems to encounter significant practical difficulties. Many vehicles on the road must remain individually owned in any event (such as vehicles serving a particular business).  With the abundant free parking already prevalent in the United States, one of the greatest cost benefits of a taxi system — eliminating the need for costly local storage — is greatly lessened. In other countries or in certain US cities, this advantage might be more appreciable, but then, a city that is sufficiently dense will, or eventually will, offer alternative transportation options that do not require storage, either. In the case of this technology, only time will tell.

-----------------------------------------------------------------------------

*Let's consider that for a second. One study has estimated that a fleet of driverless taxis could dramatically reduce car ownership, on the order of one driverless taxi replacing eleven owned vehicles. Given car ownership in the US of 800/1,000 population, this suggests a reduction in car numbers to only 72/1,000, or approximately the same number seen on the roads of Mongolia.

Now, with the savings from not having 728 owned cars, at an average of $8,800 per year per car, a surplus of $6,400,000 per 1,000 population is generated, or approximately $90,000 per remaining car just based on eliminating ownership. Even if we generously assume $30,000 per year per car to account for gas, heavier maintenance and more frequent car replacement, that still leaves $60,000 for driver wages even before accounting for any per trip fees levied on riders. Even a modest fee of, say, $2 average per ride, could generate tens of thousands of additional dollars per year. In other words, replacing private car ownership looks, at least on a back-of-the-envelope calculation, like it would be financially feasible right now, not at some uncertain date in the future.

This still underestimates the savings, however, since by eliminating private car ownership, and turning the task of piloting cars over to professional drivers, we both eliminate the need for expensive parking minimums and may reduce the economic cost of crashes (estimated at $871 billion each year nationwide, or $900/person in economic losses).

Furthermore, such a network would presumably make many local bus systems obsolete or uneconomic (though probably not regional bus or rail systems), so that spending in these areas could be transferred to supporting a public network of taxis, offsetting the additional tax that would need to be levied to fund the system. Although the retention of human drivers creates a major added expense, drivers can be expected to perform some duties (refueling/recharging/basic cleaning and maintenance) that would otherwise need to be done by separate employees, and the human brain offers a proven technological fix to many of the difficulties still encountered by driverless cars.