13 propositions about autonomous vehicles and urban transportation

It looks more and more like autonomous vehicles will be a part of our urban transportation future. There’s a lot of speculation about whether their effects will be for good or ill. While there’s a certain “techno-deterministic” character to these speculations, we’re of the view that the policy environment can play a key role in shaping the adoption of AVs in ways that support, rather than undermine the transportation system and the fabric of cities.

A rocky road for autonomous vehicles? A March 24, 2017 crash of an Uber self-driven vehicle in Tempe Arizona via REUTERS.

Our thinking is still evolving on this subject, but to start the conversation, we’ll pose 13 propositions about the nature of urban travel demand, autonomous vehicles, and what we’ll need to do to change our policies and institutions to cope with them. Given that we think that many of the persistent problems with our current transportation system stem from getting the prices wrong, we think that the way that autonomous vehicles will change the cost and price of urban transportation will be key to shaping their impacts.

  1.  Urban travel demand is highly peaked. As a rule we have plenty of capacity in our transportation system for about twenty of the twenty four hours of the day.  Because we all disproportionately tend to travel at the same times, in the morning and afternoon peaks, streets are taxed to their limits at peak hours, usually for an hour or an hour and a half in the morning, and for two and a half to three hours in the late afternoon. As Jarrett Walker observes, this is a geometry problem, single occupancy vehicles are not sufficiently space efficient that they can accommodate all travelers in peak periods in most urban environments. But it would be more accurate to call this a “space-time” problem:  we don’t have enough space at certain times.  The analyses of AV adoption and deployment routinely abstract from these issues.  The peaked nature of demand has important implications:  more economic value is associated with peak period travel than travel at other times of the day, both due to its volume, and also to the nature of demand. Demand for peak period travel is more inelastic—which is why travelers routinely endure longer travel times in peak hours rather than simply waiting and making those trips at some other hour when congestion is less and travel times are faster:  we willingly endure an extra 5 or 10 minutes in our commute traveling at the peak, when if we waited an hour or ninety minutes, we could shorten our trip by that amount of time.
  2. Parking costs shape mode choice decisions.  Where parking is “free” to end users, they are far more likely to drive. More than four-fifths of all households own automobiles. The costs of owning cars are largely fixed (depreciation, insurance), and the marginal cost of taking a trip by car is often regarded by users as largely just the incremental cost of fuel. The major additional cost to many trips, especially to urban environments is the cost of paying for car storage when the vehicle isn’t being used.  The cost of parking in city centers is a major incentive to using other modes of transportation. There is a very strong correlation  between parking costs and transit use.  In effect, parking costs act as a surrogate road pricing mechanism for trips with origins or destinations in the CBD. The advent of autonomous vehicles (AVs) will greatly reduce or entirely eliminate the cost of parking as a factor in mode choice.  Many people who would not drive to the central business district, in order to avoid parking costs will want to choose AVs.
  3. Autonomous vehicles costs will be low enough to compete against transit. The cost of AV travel may be something on the order of 30 to 50 cents per mile (and could be considerably less).  Most transit trips are less than four miles in distance.  Most transit fares are in excess of two dollars per ride.  AV’s may be cost competitive, and potentially offer much better service (i.e. point-to-point travel, less or no waiting, privacy, greater comfort, etc).  Its fair to assume that the advent of a widespread deployment of fleets of AVs will stimulate a huge demand for urban travel, both among car owning households who don’t currently drive because of parking costs (because parking will be reduced to nil), among car owning households who do commute by car (because they can avoid the cost of parking),
  4. Suburbs will be relatively poor markets for autonomous vehicles. Conversely, where parking is free, and where density is low, fleet AV service will be a far less attractive option for travelers and a far less lucrative market for fleet AV operators.  Because they don’t have to pay for parking currently, commuters don’t save this cost when paying for an AV. Also, less dense areas will by definition be “thinner” markets for car sharing, for companies this means less revenue per mile or per hour and lower utilization; for customers it means longer waits for vehicles. People who live and work in low density areas may find it more attractive to own their own vehicle.
  5. AVs will tend to concentrate in urban centers:  The markets are denser there. The technical challenges of mapping the roadway are more finite, and the cost of mapping can be spread over more trips per road mile traveled. And, importantly, they will be able to surge price in these locations. Surge pricing is possible because the demand for travel, particularly at the peak hour, is higher.  Demand is greater (more people traveling) and people attach a greater value to their travel time.  Companies will want to concentrate their fleets in places that have lots of customers, both to optimize utilization (less waiting, dead-heading) and also to maximize revenue (surge priced trips are more profitable than regular fares).
  6. The demand for peak period travel in urban centers will tend to overwhelm available road capacity, even mores than it does today.  More commuters will seek to travel by AV; AV fleet operators will concentrate their vehicles in lucrative dense locations.
  7. Surge pricing by AV operators will help equilibrate supply and demand. While AVs may only cost 30 to 50 cents to operate, surge prices in dense urban environments could be many times higher than this amount.  Operators will use dynamic pricing to ration vehicles to the highest-value users. Others who might like to travel by AV will choose other modes or times (travel by transit; pay the price of parking and drive one’s own car, wait for a cheaper AV at an off-peak time, walk, bike, etc).  AV’s will tend to fill up existing road capacity.
  8. AV fleet operators will capture a significant portion of the economic rent associated with use of the limited peak period capacity of roads.  Pricing will result in a more efficient allocation of road use among users (in a technical sense, and abstracting from distributional issues).  But the profits from the limited capacity will go to the AV fleet operators, and not the public sector, which is responsible for building and maintaining the roadway, and is typically asked to incur huge expense for additions to capacity to lessen congestion.
  9. Under current road financing policies, AVs might end up paying almost nothing for the use of the public roadway. The gasoline tax is the principal source of revenue for road construction and maintenance. Electric AVs pay nothing in most states toward road costs.  A hallmark of current transportation network companies has been their “disruptive” policies of avoiding (or shifting) the fees and taxes imposed on conventional taxis. We assume this behavior will continue.
  10. In addition, AVs will disproportionately make use of the most congested, most expensive parts of the public street and road system. Unlike typical vehicles, which as widely noted are parked 90+ percent of the time, AVs will receive much higher use, and as noted here, will tend to gravitate toward the densest markets, and due to surge pricing, will be drawn to the most congested locations. With fuel taxes, the privately owned vehicles pay the same per mile cost for road use whether they use lightly trafficked roads at off-peak times, or use congested urban roads at peak times.  As noted, parking costs effectively discourage peak use in dense locations.  And to some extent, the off-peak and low density use of cars means that some roads cross-subsidize others. Parking fees and private ownership of cars have in effect limited the ability of cars to overwhelm city streets. Both of these constraints will be largely erased by fleets of autonomous vehicles.
  11. Some regime change in road pricing is needed.  The gasoline tax won’t work for electric vehicles. Fees tied simply to energy consumed or vehicle miles traveled ignore the very different system costs imposed by travel in different places and at different times.  A VMT fee still allows private fleet operators to capture all or most of the economic rent associated with peak travel in dense urban places, and provides no added revenue to address road or transportation system capacity constraints.
  12. What we really need is surge pricing for road use. The key constraint on urban transportation system performance is peak hour capacity. Single occupancy vehicles represent a highly inefficient way to make use of very expensive peak hour capacity. Without surge pricing for roads, AV fleet operators have strong incentives to capture the economic rents associated with peak period travel, shifting costs and externalities to the public sector and non-user travelers.
  13. Surge pricing should be established before AV fleets are widely deployed.  Once deployed, AV fleet operators will have a powerful incentive to fight surge pricing because it will reallocate economic rents from them to the public sector.

Please consider this a first draft. We invite your comments, and expect to periodically revise, expand and annotate these 13 propositions.