Not another day goes by without some transit detractor talking about how driverless vehicles will obviate the need for transit and big spending on capital projects. Such statements are based on an expected improvement in travel by personal car, but they fail to account for the vast improvement in transit travel that will also result from this technology. It really comes down to three aspects of transit:

  • Frequency
  • Cost
  • Corridor capacity

Frequency is freedom

Nothing affects transit ridership like frequency. Consistent high frequency means one can depend on a transit line at any time of the day, and can get moving without delay. This brings transit closer to the utility of a personal vehicle. Frequent service allows more people to ditch the car for a trip or even permanently.

Today frequency is expensive and usually becomes available only for lines that already have high ridership. It is typically used only during rush hour to move the crowds rather than consistently across the day. When demand drops down in off-peak times agencies tend to use smaller buses–40-foot as opposed to 60-foot articulated–but no shorter. To accommodate further drops in demand agencies tend to just reduce the frequency of a route. Since around half of the operating cost of a bus is labor, the savings in fuel and maintenance wouldn’t allow for much higher frequency on a smaller vehicle.

Yet, driverless technology changes that. If labor is no longer part of the operating cost of a bus, one can scale that cost within a much greater range. One can make a decision that frequency will be constant at a vehicle every 5 minutes all day long and to keep the cost per rider constant, run ever smaller vehicles as demand goes down in off-peak times. Every 5 minutes in rush hour? Yes, still a 115-person 60-foot bus. Every 5 minutes at midnight? Probably an 8-person van or a minibus. At 3am? The same minibus waiting in a layover bay for the first passenger on the line to request it.

Overlake Transit Center
Overlake Transit Center

Overlake Transit Center is an early preview: it serves public transit (Sound Transit, Metro) and private transit (Microsoft). Microsoft provides both fixed route and on demand service and uses 5 vehicle types from vans and passenger cars (left) up to 50-foot buses (right). Moreover, Link light rail service is planned for 2023.


People cite that driverless cars will allow for cheaper ride-hailing services that can better compete with transit. Yet, as long as they are single-occupancy they won’t be any cheaper than a car you drive yourself. I am purposefully excluding cars which one shares with random other people (e.g. Uber Pool or Lyft Line) because that is in fact a form of mass transit with very small vehicles. The vehicle size on such services would only depend on corridor demand and could be scaled up within certain constraints.

So to put a number to it, by-the-minute rentals like Car2Go or ReachNow charge $0.41 to $0.49 per min. Self-driving cars would likely be equally expensive. That is not cost-competitive with mass transit (public or private) except for some very short trips or when one manages to fill an entire private car with passengers from the same group. For example, a trip from Ballard to Belltown, $2.50 on public transit, would be at least $7 in a private driverless car assuming no traffic. The cost on public transit is even lower for a frequent rider with a monthly pass.

Corridor capacity

As cities become denser, the demand for people moved per hour per street increases proportionally to the population increase. Street space, however, does not. As established in the supply and demand of street space, while a general purpose lane can move 1,440 people per hour (even accounting for our region’s average occupancy of 1.6 people per car), a bus lane can move more than 4 times as much at 6,000 people per hour! This is a capacity benefit equivalent to building 3 more lanes of road at the cost of not much more than paint. Since other road expansion options are cost-prohibitive at scale (widening requires acquiring expensive urban land and destroying neighborhoods, while elevated and underground/tunneled roads are even more expensive) it is safe to assume that we will see more space dedicated for high-occupancy vehicles.

And while bus lanes certainly enable moving more people on the same stretch of road, on many corridors it may make sense to utilize less constrained HOV lanes with a demand-based minimum number of people per vehicle. This would open the gates for mass transit with smaller vehicles – vans, minibuses or even just full cars–and also both public and private use. It’s likely that a significant portion of future mass transit will look like this. A great preview of such an arrangement are the HOV 3+ lanes on SR-520. They don’t experience congestion and serve not just Sound Transit and Metro, but also the private shuttle buses of major employers like Microsoft, Google, Facebook and also vanpools and a small proportion of cars.

So if HOV lanes are a viable capital investment, what about rail? Given all of the above, isn’t it a waste of taxpayer dollars then? One just has to look at the capacity comparison from the above-mentioned article to get a quick answer.


As grade-separated rail with 380-foot trains (e.g. like our current Link Light Rail as shown above) can move more than 5 times more people than a non-congested bus lane or even more than an HOV lane, there really is no match to getting more bang for the buck from the cost of building new right of way.

Some people mention driverless platoons of buses with no following distance making a de facto road train. However, if such a road train were to merge even just in and out of stops, it would require through traffic to make an abrupt stop due to its lengthy merge time. Even though it would be rubber tired, it would operate most safely with the same type of signaling system used for rail transit. And that would mean limiting throughput to around 60 vehicles per hour and would preclude most other uses effectively requiring dedicated right of way. In other words, such a road train would be just as expensive as a conventional rail train. So when it comes to designing the highest-demand corridors, rail is not just the best alternative today, but may remain so for a while in the future.

In conclusion, if mass transit remains cheaper than personal cars, is frequent enough to compete with its level of freedom and travels faster during rush hour, will it still be killed by driverless cars? Not by a longshot. It is more likely to experience a new renaissance with service better than ever before and will help us derive new returns even on pre-existing infrastructure investments, so we should keep funding them.

Featured image: Driverless bus in operation in the Netherlands.

We hope you loved this article. If so, please consider subscribing or donating. The Urbanist is a non-profit that depends on donations from readers like you.


  1. While the rest of the article is fine, you greatly oversimplify the capacity issue. It really depends on the vehicles as well as the roadway. TransMilenio (in Bogota) has a capacity of 35,000 per hour. They achieve such high numbers because they have dedicated roadways as well as passing lanes. Interestingly enough, Seattle had exactly the same thing with the bus tunnel (now train tunnel). Capacity was never maximized, though. Not all the buses were big and they never had level boarding. There was a form of off board payment (ride free area) but that went away a while ago (it had other problems). I don’t think it would take much, though, to achieve that sort of ridership (over 30,000 per hour).

    In contrast, our light rail line has a capacity limitation of under 20,000, before you get into problems (

    It is counter-intuitive to think that a bus system could have higher (much higher) capacity than a train system, but it really depends on each system. Buses have much smaller headways. Our trains max out at every three minutes (or six minutes for Rainier Valley). Bus headways are measured in seconds. For a given corridor, this allows a lot more buses per hour. Thus if you wanted to move everyone from Seattle to Portland, you would either need a very long train, or you would be better off with lots of buses. Unfortunately, our light rail line does not allow for really long trains. These are small trains (much smaller than, say, BART or the NYC subway).

    But that doesn’t mean that a fleet of buses, in a real world situation, will always have greater capacity than our train system. It is possible that train headways could be reduced significantly with sufficient investment. Plus there are potential problems with fleets of buses. You can get more people past a particular point (e. g. middle of the 520 bridge) in a fleet of buses all giving each other two seconds of stop time, but the buses have to stop somewhere. They also have to merge. That can be done (it is done in Bogota) but it isn’t always easy. Software systems, of course, could help with that. Long before we have automated buses (or even automated trains in Seattle) we could have software that coordinates the various bus lines through a particular corridor to maximize capacity and minimize delays.

    In the case of Seattle, I think such worries are overblown. Of course it would be nice if the trains could run more frequently, but not because of capacity, but because it is always nice if trains run frequently. If we build another bus tunnel, then it would certainly have off board payment and level boarding for the entire grade separated corridor (which could stretch from Ballard to West Seattle). Doing that would mean that the capacity of the system — the shared corridor — would probably be higher than our light rail line and it would certainly be sufficient for a city this size.

  2. Good article. I mentioned the same issue a while ago (driverless transit vehicles) and came to the same conclusion (they could be a huge boon to transit agencies). I’m sure we aren’t the only ones to come up with the same idea.

    I’ve yet to read an in depth analysis of the issues, though. I’m sure the fact that this is all just speculation is part of the reason why. It could be a very long time before fully automated vehicles (either public or private) hit the road. With transit, one of the issues is fare payment. You could have completely off board payment, but that requires a substantial investment in infrastructure (payment kiosks) as well as fare monitors. This could negate much of your savings. If you have teams of fare police taking lots of buses, then the cost could add up. But like a lot of things, even this, eventually, could be automated. Or at the very least, it could be centralized via cameras. A small crew of workers at an office could be monitoring dozens of vehicles all at the same time, and if they think someone is cheating, or worse yet, see trouble, then a smaller team checks it out. Of course, you could also just do away with fare enforcers and either not charge, or go with an honor system. Nonetheless, that seems like a minor issue.

    One of the other things I wonder about is the relative cost to maintain a vehicle. The bigger the vehicle, the higher the cost. But they also hold more. Thus you always want a nearly full vehicle. That part is fairly simple. Where it gets complicated is deciding on the vehicle and frequency (headway). Do you run a bus every six minutes, or a van every three minutes? I would guess the latter would still be cheaper (even with no labor in running the bus) but not so much cheaper as to justify the degradation in service. I’m sure there is a point where people no longer care (1 minute versus 30 seconds). In real world situations, you also get bus bunching. But achieving that number (say, two minutes) would make a huge difference in any transit system. Among other things, it would minimize the transfer penalty, which would enable much better routing (a better grid).

    That is all fixed route travel. You also have the possibility of on demand bus service, like the one that Kansas City is trying out: That could easily complement the other services. There is the real possibility of some great advances in public transportation.

  3. Interesting story 🙂

    Your Ballard-Belltown cost example: the $2.50 for transit is the current fare price? $0.42-0.49/min are the actual carshare prices? I guess transit is quite a bit subsidized, so the average price (transit costs divided by number of riders) may be perhaps twice of that. But it depends on time of day, route, etc.

    • In the 2014 Metro Service Guidelines report the D-Line to Ballard is noted as having 76.1 rides/platform hour during peak and 66.2 rides/platform hour off-peak. If we assume that’s 70/hour at $2.50 (yes, some are using passes. but go with me) we have revenue of $175/hour. The cost to operate the bus for 2014 was $142.46/hour. So the D-Line bus from downtown to Ballard could operate without subsidy and would still beat other modes because of the economies of scale of putting a large number of people in one vehicle.

      • Nice to know.

        Subsidies is quite a tricky question actually. We do not have direct user fees for road space so this can also be considered as subsidy. Or other externalities, like noise, or motorized traffic blocking pedestrian access, and so forth. This is in a way a bit like subsidy as well. I don’t think we can comprehensively assess all that, it gets simply too wide and fuzzy…

  4. This is exactly why parking lots next to suburban stations in ST3 make no sense. If you are going to build parking, the 10 minute walk shed should all be TOD. Put the parking outside of that and run autonomous buses to and from the station.

    • Agree fully. If garages are necessary politically to win the vote I hope that they can at least build ones that are convertible to housing in the future (at least 8′ high ceilings, no inclined/slanted floors, could support building more floors on top). Funny how it’s the car infrastructure that will actually be obsolete, not the transit.

      • I agree that the garages included in ST3 are a bad idea. However, an architect I know who works for ST claims that they are being designed for future retrofit to housing/commercial space. I guess we’ll see.

Comments are closed.