What Do Autonomous Vehicles Mean for Future of US Highways?

Robert Poole
By Robert Poole | February 5, 2016 | 3:21 PM EST

Attendees sit in the self-driving Mercedes-Benz F 015 concept car at the Mercedes-Benz booth at the International CES Jan. 6, 2015, in Las Vegas. (AP Photo/Jae C. Hong)

Editor’s Note: This column first appeared in Public Works Financing.

For the past several years, Americans have been bombarded with visions of a near-term future in which car ownership is a thing of the past, replaced by “mobility as a service,” thanks to fully self-driving cars. The anti-highway people love this vision, because they think it will lead to compact urban living in which people will use highways far less than they do today—and hence, that conventional projections of continuing growth in vehicle miles of travel (VMT) are all wrong. Those who are planning future highway projects have a lot riding on a realistic assessment of what autonomous vehicles (AVs) portend.

Even the Wall Street Journal’s entertaining auto reviewer, Dan Neil, has been promoting the above vision:

“Autonomy will make it possible for unmanned automobiles to be summoned, via app, to your location. And not just any passing tramp steamer, but exactly the vehicle you need for the occasion, cleaned and fueled, for as little or as long as you need. … When you’re done—poof!—it will go away.”

Leaving aside the immense inventory and logistics problems inherent in Neil’s vision, let’s look carefully at what the advent of AVs may mean for the future of U.S. highways.

To begin with, most pop-culture views of the autonomous vehicles future make it sound as if a 100 percent AV fleet is just over the horizon. In fact, full “Level 4” autonomy (no driver or driver controls) is at least 30 years away in fleet-wide terms. First, there are many unresolved problems: sufficiently high reliability of complex hardware and software, sufficiently low cost, state motor vehicle regulators’ caution, unresolved liability insurance questions, and some very difficult behavior and risk trade-offs over what rules should be programmed into the automation to deal with life-threatening situations.

For the sake of argument, let’s assume all of these get resolved in the next 10 years (which I think is very optimistic). Only then can fleet replacement begin. There are about 250 million vehicles on the road today, and normal fleet turnover is 20 years. Even if autonomous vehicles are significantly safer than conventional cars, I can’t imagine the government banning existing vehicles, and it’s not clear that everyone will want to give up being in control, or having their own vehicle. So normal turnover time should be the minimum possibility for the imagined all- (or nearly all) AV fleet. That gets us to 30 years from now. Meanwhile, we have massive highway needs that we can’t let go unaddressed, on the assumption that in the AV future we’ll need less roadway capacity.

Actually, the amount of travel people will do in various possible AV futures is a subject of much discussion. The anti-highway people’s case for less travel rests on the assumption that “mobility as a service” will lead to their preferred world of high-density, compact urban living, where people’s jobs, shopping, dining, and recreation opportunities will be much closer to their homes than they are today in suburban America. Yet there is zero evidence—in census data or anything else—suggesting that any significant demographic group, including millennials or the retired—wants to live in that imagined world.

On the other side of the issue, there is a great deal of support for the idea that an AV world, with or without large-scale individual ownership of vehicles, will lead to greater VMT than conventional state and federal Department of Transportation projections.

Consider the following: 

  • The robo-taxi version of Level 4 AVs has the potential to massively disrupt conventional mass transit, with numerous small vehicles replacing the relatively small number of buses and trains, offering true door-to-door travel; again, more VMT.
  • While high levels of automation will enable existing expressway lanes to handle greater volume without congestion, the greater throughput with less congestion will make expressways more attractive, possibly increasing VMT.
  • Many transportation researchers expect the ability of people to read, work, or play while traveling in a personal vehicle will lead to greater dispersion, not merely to suburbs but to exurbs, thereby increasing VMT.
  • Enabling those who currently do not or cannot drive to use personal vehicles will also increase VMT. 

This last point is addressed in a new report from KPMG, “The Clockspeed Dilemma: What Does It Mean for Automotive Innovation?” (November 2015). KPMG researchers convened focus groups in Atlanta, Chicago, and Denver. They talked with people about how people in their extended family might use fully autonomous vehicles once they are perfected and available. Based on their responses, KPMG made the following estimates of changes in personal miles of travel (PMT) by age group, as of 2050 (35 years from now):

      • Age 0 to 15 years: 20% increase
        • Age 16 to 24 years: 20% increase
          • Age 55 to 64 years: 9% increase
            • Age 65 to 74 years: 51% increase
              • Age 75 to 84 years: 25% increase
                • Age 85+ years: 28% increase

                    Next, they converted the increased PMT into VMT. This calculation depends on what happens to average vehicle occupancy (AVO). If AVO remains at the current national average of 1.67, this translates into another trillion annual vehicle miles of travel beyond today’s level. If, as seems more likely, cars with no need for a driver lead to lower AVO (grandma goes alone to the recreation center), annual VMT could increase to over five trillion (from today’s three trillion ballpark).

                    I have no idea if KPMG’s analysis is correct, though it seems plausible. I present it here because it reinforces what seems to be the majority view among researchers at recent national conferences. Even among metropolitan planning organization (MPO) researchers, many of whose agencies are promoting densification and transit-oriented development, the consensus seems to be that our autonomous vehicles future will lead to more VMT than envisioned in traditional DOT projections.

                    That should be reassuring news for those investing in long-term public-private partnership highway concessions. And for those still on the fence over the wisdom of such concessions, let’s remind them that all such projections involve a risk of being wrong. If the increased VMT projections on which a new highway project is based turn out to be overly optimistic, who should bear the risk of ending up with excess highway capacity? In my book, it’s less bad for sophisticated investors to bear that risk, rather than taxpayers.

                    Robert Poole is director of transportation at Reason Foundation (

                    Editor’s Note: This piece was reprinted with permission from the Reason Foundation.