Agenda: Dynamic congestion pricing for autonomous vehicles

Autonomous vehicles work. They are already replacing full-time service drivers in Uber, Lyft, and taxis.

Finally took my first Waymo ride with @andersem in Austin this week. It's real.

Part of why congestion pricing is so important:

Waymo could grow minibuses, or it could grow solo cars. Road pricing is the only way to provide both the road space & the incentive to end congestion https://t.co/pFXSVUZAdD pic.twitter.com/V4FqIRnEqH

— Alex Armlovich (@aarmlovi) November 15, 2024

Delivery vehicles might come soon. Corporate fleet vehicles. And the big jump, of course, will be when they’re available as private vehicles. It’s possible that the costs are high enough that won’t happen, or won’t happen for several decades. Let’s assume it does. What comes next?

Preferences

1. Autonomous vehicles will vastly lower the opportunity cost of time spent driving.
   • Once the transition to an all or almost-all AV automobile fleet is complete, we will be able to redesign cars to be much more comfortable. In the ideal case, they’ll be like 1^st class airplane berths or better. Being “stuck in traffic” will just mean “I’m in my mobile office” or “I’m binge-watching the 2040s Lord of the Rings remakes.”
   • Imagine having a self-driving Winnebago – for the cost of electricity, we’ll be able to take vacations anywhere on the continent without stopping work. Imagine how I’m going to spend my summers when I can shift from Quebec to Minneapolis to Moab to San Diego every day or two!!!
2. Highways will be OK to great, because AVs will cut down on vehicle spacing at high speeds.
3. Downtown streets will be a disaster. There isn’t much vehicle spacing to reduce. And AVs will be more conservative about “creative”
4. But some people (and goods) will still have places to be. If they’re stuck in traffic with all the mobile-office-workers, mobile-movie-enjoyers, and mobile nappers, they’re going to incur just as much cost as people do today when they’re late.
5. This is a recipe for extremely high congestion welfare costs. In cities, for sure. But also in desirable areas during peak seasons. Imagine the roads into the Rockies in summer if Denver office workers can get their work done while driving to their evening picnic spot
   • In case you’re tempted to think I’m exaggerating, remember that all our experience with congestion comes from people whose eyes, hands, and feet must remain attuned to the traffic, and whose vehicles are designed to keep them in safe seats.
   • AVs will enable higher highway capacity, where following distance is a major constraint. But the gains at intersections will be much smaller, and AVs will be programmed to be more cautious than human drivers on streets with pedestrians, etc. At least some places, likely including most downtown streets, will be much more congested.

Policy

1. Obviously, the solution to this problem is road pricing, which can be quite sophisticated with AV. The traditional privacy concerns about segment-specific tolling are in the rearview mirror (there’s a metaphor with a terminal diagnosis) at the point that AV tech is in control.
2. The default attitude toward regulation has been to treat parked cars with more skepticism than moving cars. That made sense when a moving car implied a driver paying a substantial opportunity cost to occupy public space. Solve for the equilibrium if downtown streets are free but downtown parking is market priced.
3. The transition from driven vehicles to AVs deserves consideration but should be secondary to thinking about post-transition equilibrium.

Politics

1. Even when pricing has large, obvious benefits, voters resist it. We still can’t toll urban interstates even though 100% of experts think it’s a good idea in at least some cases. NYC’s kinda-sorta congestion pricing highlights how the bad politics of pricing intersect with the difficult technicalities of dynamically pricing traditional cars.
   • Congestion pricing for all cars, now, would be great, but it’s not realistic.
   • And many downtown streets that are OK now will collapse under the increase in driving that AVs will bring.
2. Voters will resist tolling AVs after they have AVs.
3. So the time to begin tolling AVs is now, when they have a handful of lobbyists but no grassroots.
4. The initial price can be small – the key is to institute the architecture and expectation that AVs pay to use congested roads.

Agenda

This is research that I’m thinking about with my more tech-savvy colleagues. It’s not urgent – mass ownership of AVs would take a decade or more even if they were available for individual ownership tomorrow. But it’s important.

Researchers need to model downtown traffic with AVs. We need to think about the correct scales, in time and geography, for dynamic pricing. And we need to convince policymakers that automated vehicles should pay to use congested roads.