As driverless cars and trucks fundamentally alter how people and goods move from one place to another, how should public policies evolve? Tune in to hear this conversation led by Tony Gómez-Ibáñez, Derek C. Bok Professor of Urban Planning and Public Policy, who explores the future of state, local, and national transportation and mobility policies in the face of disruptive innovations.
Wiener Conference Calls recognize Malcolm Wiener’s role in proposing and supporting this series as well as the Wiener Center for Social Policy at Harvard Kennedy School.
Good day everyone. I am Mari Megias, assistant director of communications for alumni relations and resource development at Harvard Kennedy School, and I’m delighted to welcome you to this on-the-record Wiener Conference Call. Today we are joined by Tony Gomez-Ibañez-Ibañez, who is going to lead our conversation on public policy and autonomous vehicles. Tony is the Derek C. Bok Professor of Urban Planning and Public Policy at Harvard University, where he holds a joint appointment at the Kennedy School and at the Graduate School of Design. He has taught courses in economics, infrastructure, and transportation policy at both schools. The author or editor of books such as Essays In Transportation Economics Policy and Regulating Infrastructure, Tony has served on several government advisory commissions and been a consultant to international agencies, national, and state governments and private businesses on infrastructure and policy. We’re very fortunate that he’s joining us today. Tony.
Thank you for inviting me to talk about autonomous vehicles and transportation policy. This is a really exciting time in transportation when we’re seeing major changes. I just want to flag at the beginning that there are several other faculty members at the Kennedy School who are working on this topic, including Mark Fagan, who started an AV initiative in the Taubman Center for State and Local Government, and Henry Lee, who’s working on electric vehicles at the Mossavar-Rahmani Center for Business and Government.
I want to make three points. First, although there’s a lot of uncertainty on exactly how and when the automotive technology will evolve, it’s clear that transportation is undergoing disruptive technological change. And it’s not just with self-driving vehicles, but with some related modes, particularly electric vehicles and ride hailing, which are often seen as a package by proponents, autonomous vehicles, electric vehicles in some form of share.
As a result of these technological shifts, there are going to be some very powerful policy acts as well. And the one I’m going to spend the most time talking about is congestion in big cities, which I think will be a real problem if the technology develops as promised.
So scholars of technology often distinguish between disruptive and sustaining technological changes. Sustaining technological changes are those that support the prevailing technology and improvement, and disruptive ones are ones that undermine the existing technology. For the past century I think you could characterize the auto industry as in a sustaining technology mode. The thing that’s striking is that probably the invention of the internal combustion engine was disruptive, but since then, there’ve been thousands of small innovations that have improved the vehicle substantially but incrementally. Things like pneumatic tires, electric starters or automatic transmissions, safety glass, seatbelts. There are hundreds of these little things that have advanced it.
But now, after about 100 years, it’s time for some disruptive technology. And the disruptions are quite clear for autonomous vehicles and electric vehicles and to a lesser extent for ride sharing, which we call in our jargon, transportation network companies, or TNCs.
Anyway, TNCs upends the economics of taxis or public transit or private ownership, almost everything to do with electricity generation, transmission, storage, etc., or oil. And autonomous, obviously again, a challenge to private vehicle ownership, to taxis, to buses, to package delivery, to trucking, etc.
So there a variety of different levels of automation are possible. And the Society of Automotive Engineers has a six-point scale on the degree of autonomy that a vehicle has, where zero is no automation and five is complete automation, no role for the driver. In between our level one is driver assistance, what we have already on many vehicles like speed controls, breaking, lane adherence, and parking guidance. And two and three reduce the driver’s role. Four is where it really gets interesting. There’s a driver on standby to intervene in the event of an emergency, maybe is notified by the vehicle. And five is complete automation. It’s four and five that are really the key issues.
I mean, I can’t imagine, actually, that four will work because imagine being woken up out of a nap or whatever to deal with an emergency, it would take quite an alert person.
Anyway, there’s lots of uncertainty about how and when the technology will develop. Will level four and five be available in the next 10 or 20 years? How quickly will it be adopted? And I’m especially worried about the problems of having a mixed fleet of some with high levels of automation and some without, and which technologies are going to be used. It’s going to be a mixture of three things, of sensors and computers that are on the vehicle, of signals from infrastructure to the vehicle or the conditions of the road ahead, and signals from other vehicles about driver’s intentions or conditions on the road ahead. Whether they’re going to turn left right in front of or not.
And the manufacturers obviously prefer onboard and so the state and local governments, because I don’t want to have to re-equip all their infrastructure with special, but it’s probably going to be a combination of the three.
So the next question is, what the policy effect are of these. And there are two policy effects that I think are undisputedly beneficial. One of them is traffic safety. We have about, in very round numbers, 40,000 traffic fatalities a year, and 90 to 95 percent of those are driver error, often drunk driving or speeding. So if you could reduce driver error by getting rid of the driver entirely, you might reduce some of these. Now, of course, there’s automated vehicle error too as the killing of the pedestrian in Arizona a few months ago showed.
The second area or beneficial policy consequence is environmental. Transportation is responsible for roughly 30 percent of greenhouse gas emissions. And electric vehicles hold the possibility of decarbonizing that sector by substituting clean energy and solar, wind, hydro etc. There are several challenges to this. One is getting enough clean energy. And second is that the sales of Teslas and Volts and so on currently depends greatly on the tax credits that are offered by the federal and state and local governments. And it’s not clear those are affordable or necessarily sensible.
But luckily, battery costs are coming down fairly steadily. And then the final problem or challenge is charging tech stations and technology. But I think in traffic safety and in cleaning up some of the greenhouse gases is likely to be beneficial. There’s a third effect, which is whether you regarded this as beneficial or not, it really depends. And that’s important. The Department of Labor estimates that there are five million professional drivers in the United States, of which about a quarter of a million are taxi drivers and 3.5 million are truck drivers. And you can well imagine many of them losing their jobs. Whether you regard that as good or bad, depends upon whether you think that there’s something productive for them to do and that’s, I think, an open issue that will depend much on how rapidly the change comes.
There’s another condition ... another circumstance or policy outcome that I think that is going to be undoubtedly harmful, and that’s the impact on congestion. And this is a simple story in many ways. I mean, AVs eliminate the driver. That means for chauffeured modes, for modes where you have a driver like taxi, or bus, or truck. You save approximately 40 or 50 percent of the gross. In a big city. In bus transit in a company for example, about 42 percent of the total operating costs are for the driver labor, and that would be eliminated.
And with “do-it-yourself” modes like our ordinary private automobile, you save on the inconvenience and anxiety of commuting or driving. Indeed some people talk about the possibility of commuting from Stockbridge, getting in your car and taking a nap as the car drives you into Boston. But the key here is to understand that reduced costs whether of chauffeured modes or of private modes, “do-it-yourself” modes, rather, means increased use, right? Just the fundamental of economics. You drop the price of something, you increase its use.
Now, proponents of autonomous vehicles point out that they may be able to increase capacity as well by driving closer together, or closer at intersections. But I think that that capacity increase will unlikely be enough to offset the increased demand for travel that it needs to provoke. And this is especially true given that the proponents of AVs is one of a bunch of other policies in there. One is, they assume that their AVs are going to be electric. Well, if you think about the cost structure of electric cars, they’re high on fixed costs and low on variable costs relative to a gasoline-powered vehicle. Their maintenance is very easy because the motors are very simple, for example.
So what you’re going to do is increase again the incentive of people to drive more because it’s going to be a low-marginal-cost-car and shared vehicles are very important. Proponents argue that if we share vehicles, we’re going to have much higher vehicle productivity and a lot of parking spaces liberated because most parking spaces are only occupied 5 percent of the time or less and most cars are only driven for 5 percent of the time or more.
And this is a big advantage, particularly cited by some urban planners who covet the land from the parking lots, as well as they should. But I think while it increases the utilization of vehicles and parking spots and hastens the adoption of the technology, it reduces marginal cost and increases congestion even further. And if you’re going to spread the cost of the car or parking spots over many users, it’s obviously another marginal cost reduction. And we’ve tried also to increase vehicle sharing by having Zipcar and other devices. And that has not worked in part because people want to have their own car with a car seat if they have kids and whatever other features they want.
What you really want if you’re going to reduce the congestion effects of AV, it’s not shared vehicles, but shared rides. You need to pool. And that too, we’ve been trying for decades to do with carpooling incentives since the energy crisis of the early 70s and not very successfully. I mean, we’re trying again with Uber pool and Lyft.
And the TNCs, Uber and Lyft, may be better at matching than our old carpool matching programs, which tended to be on the bulletin board around the cooler of the office. But I think what you’re going to need is some kind of dramatic response to this congestion problem without ... to prevent it from getting out of hand, particularly in big cities. And I think it’s likely to take the form of congestion pricing. And for the charge to grow it depends upon how congested the road is that you’re on.
I might speak a little bit about the TNCs if I’ve got the time. The TNCs like Lyft and Uber are really relevant in this discussion. You know, like the AVs, the TNCs are less than the current taxis in price, right? They’re 20 to 30 percent, usually less. And that’s going to increase their use. And the question then becomes, whether their ... the key question that everybody wants to know is whether they are substitutes for the existing public transit services. Now, will they drive them out of business or really make them even stronger? What people talk about is, or optimists talk about is using the TNCs to offer the last-mile and the first-mile service.
But I think more generally, it’s to exploit their flexibility to fill in routes or sections of town or hours of the day where there’s little service because the density is too low and the cost is too high. So I’ve got to fill in the holes in the net. And when we’re talking about substitute, we’re talking about often cannibalizing or skimming the cream of the transit system. Now, there isn’t a lot of cream, financial at least, in the transit systems in the United States. But what they’re more generally talking about is replacing existing service where the incumbent is enjoying some economies of traffic density with a slightly superior service that does not exploit the economies of traffic density.
Anyway, there are two sources of information about whether or not the TNCs are compliments or substitutes to public transit. One of them is surveys, which asked TNC users what mode they would have taken if TNC had not been available. And there’s an ICE University of California Davis survey of the literature, which shows that it’s pretty consistent that TNCs add to traffic and congestion. They report their summary of all the surveys that 39 percent would have driven or carpooled or taken a taxi without the TNC, but 61 percent would have taken fewer trips, transit, or biked or walked. So you’re actually increasing the amount of traffic rather than reducing it.
I don’t really trust these surveys that much, in part because they’re stated preference rather than actual, revealed preference. But in this result I’ve been very interested in some econometric work done by Professor Hall at the University of Toronto, who estimates the effects of the entry of Uber on transit ridership in the United States and finds that Uber ... most compliment systems that are small relative to the metropolitan population is where they can fill the holes that they’re going to be used.
Finally, I think the future of TNCs is another fascinating topic. You know, will this ride hail system completely replace the traditional street-hail taxis? I think it’s unlikely to do so. I think that the taxi and the TNC are going to merge together because that differential of 20 and 30 percent, it’s hard to see how it’s going to be sustained. It’s not going to be sustained by investors. Maybe the investors are not going to be willing to subsidize it as heavily as they have now. It’s not going to be sustained by having to pay off by medallions since the medallions are being eliminated. It shouldn’t be sustained through labor or savings, because they’re both competing for drivers in the same labor market and will have to pay comfortable wages, and that the apps, the car matching apps or ride matching apps which are the real contribution of this technology to us, don’t seem to be compatible or incomparable rather, among the different TNCs. You don’t hear Uber complaining, Uber drivers complaining that this software is better.
And I take it that means that the softwares are not unique and they’re pretty competitive. There is one factor that may allow the TNCs to collect a sustained surcharge. And that is what we call network economies, that the competition for riders and drivers in that competition, the advantage goes to the largest network. If you’re a driver, you want to join the largest network because that will get you the most rides quickly. And if you’re a rider, you want to join the largest network for the same reason.
The competition for riders may be effective if people are willing to shop around. Here’s one of the things you can do obviously, is if Uber’s has surge pricing, you can go get a quote from Lyft. And I don’t know how many people are willing to do that. I’m actually ... I don’t do it. I confess I only have Uber.
The competition for drivers may be more problematic because they can’t switch back and forth between providers with the smooth ease that the passengers can. But you could always require that there be no exclusive contracts so that drivers could have both an Uber and a Lyft app in their car. But in any case that’s all I have to say about the fascinating issues of automated vehicles and I look forward to your questions and comments.
Q: How should municipalities think about risk with regard to transportation network companies, or TNCs, or with autonomous vehicles, especially amplified by today’s social media as they roll out new policy and regulatory structures? If we could talk a little bit about the policies around that.
Well, I think that the rollout of AVs is going to be dependent very much on traffic safety issues and accidents will have a big effect in how they’re handled. I would hope that people will keep in mind that we kill 40,000 people with regular vehicles. And that’s a terrible thing. Killing no one is probably an impossibility, but we should be able to reduce that number by a lot. Anyway, it’s going to be very interesting to see how that plays out.
Q: I wanted to address one of your points about the effects of automation and TNCs on the marginal cost of travel. I believe you are arguing that TNCs would further drive down the marginal cost of travel and increase travel demand. But, as you probably know, there’s a prevailing argument that TNCs will essentially force the traveler to see the full cost, full average cost of travel in their fare. Like TNCs now currently cost over $1 per mile, whereas a private traveler views the marginal cost somewhat mistakenly, though, as essentially fuel costs, which is on the order of 10 cents. So the argument is, the TNCs will cause people to see the ... as their marginal cost, the full average cost of travel and that might actually deter some travel. What do you think of that argument?
I think that that has a lot of merit. I think that their econometric studies out of Toronto are better at answering that trade off and suggest that the arrival of Uber can cause an increase in traffic in some places and in a decrease in transit in others depending upon how dense the existing transit service is, whether there’s opportunities to fill the holes or not. I think you’re right about the difference between the average cost and the marginal cost of public ... of TNCs versus private driving. But it hasn’t shown up much in the statistics yet.
Q: Because we expect driverless vehicles at whatever level of technology to be geo-fenced in their first several years, does that say that autonomous vehicles at the beginning will all be robo taxis and they will not be very much private ownership of these vehicles, do you think?
I don’t know. It sounds plausible that there would be only brought by people who could use them intensely because they’re going to be expensive at least initially. And if they are geo-fenced, then there may be only specialized users who can use them as intensely enough to make some competitive. So it sounds plausible.
Q: It appears that the industry is driving the technology but localities and consumers may want choices including existing types of vehicles. What choices will consumers have, people with special needs and an aging population as examples of consumers who may not want autonomous vehicles and if given no choices may create unintended social consequences?
Well, I don’t think that people are going to have no choices. I think the idea that we would share vehicle ownerships because vehicles could be owned in fleets that were operated and maintained by the big car manufacturers doesn’t strike me as realistic, doesn’t strike me as respecting kind of democratic norms almost. No. And there are legitimate reasons why people might want a different car than everybody else. Or by one of their ... I had one woman tell me that she doesn’t smoke and she hates being in rental cars where she is convinced that ... And another person told me that they had three kids and they couldn’t imagine anybody wanting to be in their car after the three kids had been in their car. So, you know, there are people who like to camp and there ... People have lots of different reasons for wanting that. And in theory a fleet operator could offer a menu of vehicles that would satisfy our needs but that would be hard. And I just don’t think that there are that great advantages to it. In my experience I don’t know. Do you follow the experience of Zipcar, which was created about 10 or 15 years ago, I think, and was a car ownership program where you bought a share in essentially a carpool. But they never took off it only remained I think one or 2 percent of the people that were car owners in the metropolitan areas where it was available. And I think that speaks to the top because of people’s individual preferences that was obviously ... are cheaper to own a Zipcar or sharing a Zipcar or invest on the car itself and not very many people chose to do it.
Q: As you’re probably aware, in a number of major cities, congestion pricing is being actively explored. And I was wondering if you had any thoughts on how autonomous vehicles should be treated in such regimes? Would they be potentially subject to preferential treatment?
I don’t know why we want to subject them to preferential treatment. I think that the idea of congestion pricing is to let you have each vehicle responsible for the congestion that it causes. And so if AVs cause more congestion they would be charged more. But I don’t see any reason why AVs would cause more congestion per vehicle mile if they were traveling at the same time and at the same place as an ordinary vehicle. But I thought you were starting to ask Uber congestion price and the politics of it which are at least I think are more ... are difficult at least. But I’m encouraged by the fact that that New York is seriously considering it now again. And think that it’s places like New York and Boston, Boston wants to remain competitive as a location for industry and offices, that will end up leading the way and largely forced by many surplus of automated vehicles.
Q: The question I had was around, whether it’s autonomous vehicle specifically or sort of these different mobility modes that you’ve been discussing, it seems to me that a public-private partnership model is really something that we could use to, as far as I’m concerned, break a logjam of how we bring this technology to market and how we bring it to market safely. Currently there really is no rigorous testing, although we’re working on that in industry. Do you have any suggestions on how we could get the federal government, state and local, private sector, and maybe research sectors together to look at this and do more of a crowdsourcing that could speed this to market. And as you mentioned, make significant, not just congestion lanes but safety lanes and make a dent in those 40,000 deaths you talked about.
I don’t know how much of a role I want the government to play at the moment in this business. I think the technology is still so new and developing that it’s almost better to have a variety of different companies, and different types of companies, approach it and see what they can do before we make too much of a commitment of government resources, at least for the moment.
Q: Tony, I have a question regarding something you had mentioned earlier about the use of these autonomous vehicles as part of a last mile or first mile of service. For example, perhaps, connected with networks that run from large transit rail stations and that sort of thing. Is there any roadmap that you’ve seen within North America to have these sorts of systems put in place where you’d have, for example, dedicated lanes for autonomous vehicles to take passengers short distances, let’s say to their offices or similar types of establishments.
You know, you haven’t seen it really on the last ... on a downtown mile, but you have seen people experimenting with access on the suburban end of these services. There’s been, I think about a dozen attempts by public transit agencies to offer various kinds of last mile, first mile service in the suburbs in conjunction with Uber or Lyft. And some of them have been reasonably successful and some of them not. And it’s not clear to me why. I don’t know enough about the individual demonstration projects to tell you, but I would look there.
Q: I had a question about the timing for wider use of autonomous vehicles. And you had mentioned earlier that you thought driverless trucks would come first, I think because of the interstate system and the fact that most miles are driven on the interstate. Can you talk about the timing of when those driverless trucks could affect the 3.5 million truck drivers employed in the trucking industry over the years?
I’m not a technology expert. And so I am not confident in any forecast that I would make about when a technology is going to be ready and when it’s not. My guess, though, is that the trucking on something like the interstate is by far the farthest along. And my guess is that they could probably do it now. It wouldn’t be ... They’ve run some experiments where they’ve had platoons of trucks. There are some practical problems involved and people are going to be scared. Imagine driving along the highway and seeing a platoon of driverless trucks. But I think it could happen fairly soon, maybe within the next 10 years.
Q: I’m from Mexico, it’s been argued that the younger generations are the early adopters of technology in general. But I read at the Economist this week that the elderly might be the early adopters this time for various reasons. One of them is that they live in close communities where their urban planning is easier and the speed limit is slower. What are your thoughts on this?
Well, as a senior citizen, I just want to tell you that I use Uber a lot for the very reasons that you’re talking about. I don’t to want to tell you too much about my situation, but with my cataracts it’s hard to drive at night therefore it’s doing an excellent job. And I think that there are millions of people that are like me and this would be an important market.
Q: What do you think the impact will be on local and regional transit agencies? And what should they be doing now to prepare for the changes that are yet to come?
I think the impact is going to be tremendous. And that they ought to be thinking about the economics of compliments and substitutes, right? Where they can ... where there are holes in their network that they can take advantage of the flexibility that TNCs or AVs provide and whether there are parts of the network where they’re not enjoying very many economies of scale that they could give up to the TNCs and have them operate at a lower cost and vice versa. You know, where there are segments of service where they are enjoying economies of scale and where they don’t want to give it up too. But they ought to be examining their services with those ideas than not.
Q: Can you address the role that local government can play in this whole new arena?
Well, local government has always been responsible for the regulation of the streets for which the AVs will have to use just like ... and the TNCs have to use. And so their policy about street use and about parking, on-street parking and use are going to be critical to give them important leverage over this. I think the issue of how a street curve is used is an issue that we’ve just woken up to. For those of you who are not frequenting Harvard Square anymore, there’s an epidemic of double parking by Uber and Lyft cars in the Square. And that’s obviously an issue for the city of Cambridge too. Same with electric scooters and bikes, and we’re sure we going to ... We can’t provide the priority to everybody because they’re in ... So I think we’re going to have to make some choices about what we provide priority to. And that’s another job for the local government.