David Levinson is Professor of Transport Engineering, School of Civil engineering at the University of Sydney. Details of his many published articles, and books, can be found on his blog, Transportist.
Good morning, David. Could you tell me a little bit about where you are at the moment and what work you’re doing?
I’m currently at the University of Sydney in the School of Civil Engineering, directing the TransportLab research program. My research reaches into a number of different areas. I have a program working on measuring and advancing the theory of accessibility, how many things you can reach in a given amount of time. For instance, how many jobs you can reach within 30 minutes.
I have students looking at understanding the co-evolution of transport networks and land use over time, using Sydney as a case study. How were trams deployed? Did the trams lead or follow real estate development? And what does that say about transport and land use co-evolution going forward? I’ve done work on this field in the past in other cities, including London and Minneapolis, and since I’m here it was only natural we would look at Sydney.
One student is examining game theory models of lane changing, in particular, can we design socially-beneficial lane changing algorithms that might be used in autonomous vehicles? Having vehicles change lanes when it would save other people time and not change lanes when it would cost other people time.
I also have a student evaluating benefit cost analysis processes. The idea is that while a lot of benefit cost analyses compare alternatives, we don’t know how different are those alternatives? How accurate was the benefit-cost analysis in the first place? And if the accuracy is less than the difference between the alternatives, maybe it wasn’t as important as people think it was.
But if it turns out that the difference in the predicted benefits and costs, compared with what actually happened was relatively small, then benefit-cost analysis is important in determining the best outcome.
Sounds like you’ve got a bit going on and so has your department and students! Moving back in time, what set you on the path of a career in transport?
I grew up in the city of Columbia, Maryland, which is between Baltimore and Washington. Columbia is a planned community. When I was very young, in third grade, we were exposed to the ideas of city planning and I became interested in how to design cities. And how to design efficient cities. Of course, the core of that has to do with how to things relate to each other in space and how are they connected, which gets you into transport.
I was thinking about that career for a long time, and then in middle school, and especially high school, I became interested in computers. That interest saw me start at Georgia Tech as an electrical engineering student. But I decided that wasn’t really as interesting as I might like, so I made the switch to civil engineering, with the idea of getting into transport planning, which I then ultimately did.
My first job out of Georgia Tech was building travel demand models, which combined computers and transport. That was in Montgomery County, Maryland, outside of Washington, DC, and saw me as part of a team building a regional planning model to understand flows between places. This was an area that really grabbed my interest, and I went to graduate school, doing my PhD at Berkeley.
My study was in the area of transport economics, because I thought that the main problem was really the problem of demand. We don’t price streets and roads, especially in the United States, certainly not at what they cost, and so people over-consume roads. This is still true, people over-consume roads in general. The motorways in Australia might be tolled at too high a price here, but the other streets are tolled too low, and the tolls don’t vary by time-of-day.
I also undertook a project evaluating California’s high-speed rail project, which was just an idea at the time. Then I took my first academic permanent position at the University of Minnesota, where I was until 2016 before moving to Sydney in 2017.
In retrospect how did your hometown of Columbia fare in terms of transport planning?
Columbia is, in a sense, too suburban according to current planning theories, and the density is too low to enable effective public transport between places. When it was planned back in the 1960s there was to be a minibus circulator system connecting all the neighbourhoods to the village centres, and then with the Town Center. They had planned a separate right-of-way for this network, but built very little of it, and then abandoned it (some of it was later converted to bike paths). There are suburban buses within Columbia, and a few to Baltimore and Washington, but with really low frequencies.
From a public transport perspective, it didn’t work out really well at all because you need to have a critical mass of demand in order to have public transport work. One of the things about Sydney compared to US cities of a similar size, is that public transport’s really important here. The commute mode share for public transit in Sydney is 21% or so while in similarly sized US cities it is much lower, (well below 10%) and that’s interesting.
So what’s different about Sydney compared to US cities? Obviously it has a higher density of activity. It was built around trains. Although a lot of US cities were built around trains, they then abandoned them, but Sydney did not. Sydney did abandon its trams and replaced them with buses, but buses here still carry a lot of people, whereas in the US buses are seen as a second-class form of transport. Columbia being a third- or fourth-ring suburb didn’t have the population density to be able to support public transport. That resulted in a lot of people using cars and although it has an extensive off-road bike path network, the cycling community’s still small.
From a transport perspective Columbia wasn’t terribly successful in being innovative – the minibus network failed – or being environmentally sound with its reliance on the auto. Though it did get me interested in planning, I wouldn’t want to emulate Columbia. It’s not an icon of good planning in the modern sense of how we would try to design places. It’s a pleasant enough place to live I suppose, but it’s not a place that you can get by easily without a car.
Amidst that answer you mentioned road cost, that invariably the cost of roads is too high or too low. How do you think we get to a Goldilocks level of it being just right?
It’s going to be an iterative process, but I think as automated vehicles and electric vehicles in particular are deployed it will shake up the road funding system. That’s because electric vehicles don’t pay tax on motor fuel use. There ought to be a separate pricing regime for vehicles that don’t use petrol, which would be that they pay per kilometre. They would pay a higher price in peak times and receive a discount in the off-peak, to recover the costs that they impose in terms of congestion, infrastructure wear and tear, and their environmental costs. EVs don’t impose as much environmental damage as internal combustion engine vehicles, but it is not zero.
Autonomous vehicles present an interesting problem as well because today for a car to drive around, the driver is at least spending their own time. They don’t generally drive around circling the block for the hell of it, because they would be wasting their own time and wasting fuel. In an autonomous vehicle, you can avoid parking by ordering the car to just go around and circle the block and not waste your own time doing that. This however wastes everybody else’s time because you’re congesting the roads.
The solution will impose some price for driving. Either it would be illegal to circle the block empty, with an associated high fine associated, or better and simpler, there’s a per kilometre charge for driving in cities. And it’s a higher charge in cities than it is in places with less congestion. That would discourage that kind of behaviour so that instead of driving around using the road space as a way of avoiding paying for parking, people, if they wanted to keep their car near them, would find that paying for parking was the less expensive solution.
I think in a sense these technologies are going to force a change in transport financing one way or another. It would be good to get ahead of it. Western civilisation is not really very good at getting ahead of problems and generally only reacts to crises rather than acting in anticipation. This is unfortunate, but we haven’t really figured out a solution to this political problem. Politicians have to (figuratively) be beaten over the head with a problem until they feel the need to actually resolve it, rather than say, ‘Well this is going to become a problem.’ Road financing problems are not secrets, everyone in the field can see it coming, we should be able to work on a solution ahead of time.
Like this interview? See the rest of our Meet Smart Mobility Experts interviews.
Human nature and all that. Just to veer away from reality for just a second, if you were given a very large budget and no small timeframe in which to do so, what project would you like to see undertaken that would have an appreciable impact on either traffic or however we get around, for ways of improving it?
We could do a lot if we implemented road pricing for instance, right? We could reduce congestion more than with any other technology. And the toll wouldn’t have to be that high because there’s a lot of traffic, even in peak hours that has some flexibility in when it can travel. With a small incentive, a lot of that travel would move to off-peak times when the roads aren’t already saturated. There are research projects that examine this, but what we really need is the political will.
And if you gave me enough money, I could probably buy enough politicians to do that. [That’s sarcasm]. We don’t have the political will to actually implement road pricing – the fact it doesn’t already exist argues it is going to be politically difficult. Politicians aren’t re-elected without understanding what’s politically popular and what’s politically unpopular. And obviously there’s going to be resistance to something like that.
So the next best solution is to phase pricing in over time, with electric vehicles, as more and more EVs get adopted. EV deployment is going to happen with or without government subsidies and policies, though it might be faster with subsidies. EVs are over the long run going to be a better technology than the internal combustion engine. Renewable power prices, battery prices, and the cost of manufacturing EVs continue to fall as EVs are manufactured at larger scale.
This will inevitably happen, but with policy help, and with incentives, it could happen sooner. So that’s one class of things, if we’re talking about pricing.
Getting autonomous vehicles deployed sooner rather than later is something else we could talk about. I think that’s important primarily for safety reasons, although also for capacity reasons on certain facilities.
Indeed. And the second part of the hypothetical is this time, short time limit, small budget, but would make a big change. What would you attack?
The lowest hanging fruit is not particularly technology-related, is how we allocate road space. We give far too much road space to automobiles compared to buses. We should have more bus lanes than we do, given the number of people who use buses. The objective should be moving people not cars. Similarly, we should have many more lanes for bicycles and other micromobility devices so that they can ride at speeds that are faster than walking, so they’re off the footpath, but slower than motorised vehicles. And the bicycle riders would be safer because they’d be in protected lanes with appropriate traffic signal treatments at intersections so that they get priority and their own traffic light.
Reallocating road space is something that we could do now, we don’t need any special technology to do. Other countries do this, but Australia doesn’t. The few local responses to COVID-19 show both its feasibility and benefits, but are insufficient. The United States doesn’t do this on a large scale either.
You mentioned in part of that answer, micromobility and the need to change the makeup of footpath and road space. How do you think, in an ideal situation, space for micromobility should be implemented?
e-bikes in many ways, should sell themselves, but people will only be willing to ride them if they feel safe riding them. And everything that we do in Sydney at least, but in many other Australian cities as well, is very hostile to the cyclist. Cars ride closer than they should, and are aggressive and hostile towards cyclists. There’s an attitude that cyclists are not fully human. There’s an interesting paper on that out of Monash University, that some people dehumanise bicyclists, treat them badly, cut them off, which makes on-road cycling an unsafe and generally unpleasant experience.
And rather than being cyclists themselves some of the time, most people aren’t, most people haven’t ridden bicycles since they were children. Generally there was a lot more bike riding when today’s adults were children than there is today. And what’s the reason for that? And one of the reasons for that is conditions were better 30, 40, 50, 60 years ago to ride bicycles than they are today. There are a number of relevant social trends, there were fewer cars on the road, schools were nearer homes, children were given more independence, there was no internet enticing kids to stay inside, and so on.
So we should promote more bike paths, protected bike lanes and the like, which will make it more pleasant to ride a bicycle. And the more pleasant it is to do something, the more people will do that. But we’re barely moving in the right direction. The City of Sydney has plans to increase the number of bike lanes, but it’s really pretty modest compared to what it ought to be, compared to say, what’s being done in Northern Europe, the UK, and even some US cities like Minneapolis or Portland. If you’re serious about promoting cycling, you have to show a commitment. The commitment is dedicating space so that cyclists aren’t fighting with cars for road space. As with buses, that’s something that we ought to be setting space aside for.
Now how do we get more people to use e-bikes? Well, if the environment for riding is better, people will be more likely to use them. The fines for using them inappropriately are excessively high. So for instance, if you are riding a bicycle without a helmet, there’s a large fine, even if you’re not doing it on the road, even if you’re riding on a bike path, on a trail, at a low speed, there’s still the prospect of a fine for riding without a helmet. Now, there may or may not be a safety benefit to helmets. There’s a large debate in the literature. Clearly if you’re dropped on your head, wearing a helmet is better. But if you’re wearing a helmet, you’re more likely to be dropped on your head, both because cyclists will feel more impervious to harm because they’re wearing the helmets and be more aggressive when riding, but also because cars will feel that people who are wearing helmets can handle a crash more readily and will be more aggressive towards them. And so some car drivers behave badly. Seeing someone in a helmet and thinking, ‘Well, they’re geared up for battle, then they can do battle.’ Rather than seeing them as a person, the helmet encourages drivers to perceive cyclists as an object. An impediment to their road use.
So it’s one thing to say that if you’re riding in traffic, having a helmet might reduce your brain injury risk, which it might. If you’re in a crash it probably does. But if it increases the likelihood of a crash, that’s not as obvious. And since the more punitive helmet laws have been imposed, there have been fewer cyclists in New South Wales than there were. While public policies say they want to increase cycling and active transport, they are also simultaneously working against it.
And I think that’s a problem. If you have to wear a uniform to commute to work, that’s a problem. You should be able to commute to work in regular work clothes, the way people do in other parts of the world. You shouldn’t have to get geared-up and ride at bicycle racing speeds. It’s not the Tour de France, it’s a normal commute trip.
Now, electric scooters have been shown to have high crash risks. And part of that is you have people who are inexperienced riders. Certainly alcohol is related to that, but you also don’t have great facilities for scootering, and if you had bike lanes, the environment in which to use a scooter would be better.
The same applies to skateboards and electric skateboards, which we’re seeing more of on the roads of late. There’s a lot of risk associated with these modes of transport. If they had a separated lane to use, where they were less likely to be hit by an automobile, that might be safer and might encourage more use.
That sounds like it’s a bigger job than you would indicate. How would you do that in a small way, with a small budget?
Well, I don’t think there’s a lot of budget involved. It’s just some paint on the road and putting some concrete barriers up. You can make it as complicated as you want, but if you look at, say, what New York city did under Janette Sadiq Khan, who was the transport commissioner there, they basically did a lot of trials. They closed Broadway to traffic as a trial and they put up some temporary barriers to see how it worked, and it worked fine! They eventually put in more permanent, nicer looking barriers. So put up planters and redirect traffic, take away a lane … we take away lanes for construction all the time. It’s easy to do technically and it’s just a question of prioritisation.
Traffic signal timings are another one. I don’t want to say it’s just flicking a switch in a control centre, but it’s essentially just flicking a switch in a control centre, to give more time for pedestrians to cross the street, to make a traffic signal cycle times shorter so that people don’t have to wait as long at intersections and so that people have more than six seconds to cross the street. We should be counting pedestrians automatically at intersections. This is something that the technology is available to do, it’s not very expensive and that should be figured into the traffic signal timing algorithms, which again, is not technically very difficult to do — other places do it. It’s basically a knowledge transfer problem, and implementing these small changes creates an environment that’s better for pedestrians.
Right now, pedestrians waste about a quarter of their time stopped at traffic signals when they’re going through a city. And compared to a hundred years ago when they didn’t waste any time stopped at traffic signals because there weren’t any, the pedestrian conditions have become significantly worse. We have redesigned our road system to optimise for automobiles and automobile drivers, rather than people on foot. If we want more people to be walking and more people to be on bicycles, we need to optimise the system for people walking and people on bicycles. And this includes road rules, traffic signal timings, allocation of road space, and a number of other matters.
You’re right. All of those things are simply a change in algorithm or clicking a switch. I guess the real stopper is in making a decision.
Yes. Someone has to be decisive and someone has to have the will to do it. And I think no one in power right now has the will to do that. Hopefully soon someone will have both the power and the will to do. But it’s a decision, making conditions better for people on foot, on bike, and in buses may make conditions worse for drivers. Other cities can serve as examples.
Paris, one of the world’s great cities, is looking towards banning automobiles from the entire city centre. This is a major step forward. There are not that many automobiles in the Sydney CBD as it is now, but why is so much space given over to so few automobiles to begin with?
And not just the CBD, it’s everywhere around greater Sydney. Not to say that people can walk from home to work or walk from transit to their destinations everywhere they are, but they can do it in a lot more places than do it now and we should be working towards maximising rather than minimising walking.
Okay, back to you and the work you’ve done now. What projects so far have you been most proud of?
I like the accessibility work I’ve done. Back in the US we built the Accessibility Observatory and we managed to measure accessibility for every one of the more than 11 million census blocks in the United States. Counting how many jobs can you reach in 10, 20, 30, 40, 50, 60 minutes for every point in the United States by auto, by public transport, by walking and by cycling. I think that’s a pretty nice piece of work. And obviously, I had students and researchers working on that. I really like my transport-land use co-evolution work that we’ve done.
I did a case in London when I was on sabbatical there, and followed with work in the Twin Cities to try and understand the causality of this process. Does investment in infrastructure automatically drive land development or does land development automatically attract infrastructure, and which if either of those is the dominant feature? It’s not settled, but I think it’s a really interesting question.
I developed game theory models of congestion pricing, which I think are pretty interesting.
Overall I like my work. I like most of the projects I’ve done. I’ve done a lot of papers so I don’t sit down and rank them in terms of which ones are most interesting and so on. Obviously you’re fonder of some than others. But the areas that I’m still interested in are relate to questions I have worked on previously. I’m still interested in game theory, I’m still interested in land use-transport co-evolution, I’m still interested in accessibility. We’re still doing work in those areas. I think technology deployment is also an interesting question. How quickly do we adopt technologies?
As an example, in what year would there be a 50% market share in new cars for autonomous vehicles? We don’t know the answer to that question, but it’s pretty important, and at some point we will reach that level. But do we reach that level in 2025, 2030, 2040, 2050? There’s a lot of debate about that, but it affects policy now, because if autonomous vehicles provide as much capacity improvement as the proponents claim, then we should stop building roads.
If, as proponents claim, AVs are that much more efficient, if they can follow that much more closely, and adhere to narrower lanes, we can start taking away lanes and unbuilding roads. The sooner it happens, the more we can do that. But while long-run plans that are written by agencies around the world mention or acknowledge that autonomous vehicles are coming, they assume that nothing will change in response to autonomous vehicles. I think that’s a mistake.
Autonomous vehicles will not only be more efficient from a capacity perspective, they might also increase the amount of travel. Because all else equal, if it’s easier to travel, then you’ll travel more. This is something we understand very well conceptually and estimates are we might travel 50% longer if we’re not actually actively driving, and we might send cars on all sorts of new trips.
People today who are unable to drive (e.g. children, the disabled, the elderly, people with suspended licenses) would be able to be passengers in these cars. And if these cars are inexpensive (cars in a world with all AVs can be lighter and smaller, and EVs have fewer mechanical parts and will be simpler to construct) and they don’t require a driver, we’re going to wind up with potentially a lot more travel. This might offset any capacity benefits that we get out of them, which might mean that congestion will go up, which again gets back to the question of how do we manage road space and how do we price roads, which is something we ought to be doing more aggressively already. But it’s going to become more and more apparent as we have these driverless vehicles going around.
But on the other hand, maybe there’s a substitution, maybe people don’t go shopping anymore because we’ve all moved over to a delivery-based economy. And if we can get the logistics for that right, — instead of going out and coming back, then we have a truck that goes down the street once an hour delivering everyone’s latest order or whatever — that should reduce the total amount of travel.
I think there’s a lot about the outcome of these technology trends that are unclear, but the policy-side really should be paying attention, these questions are yet to really enter the realm of public policy or long-term planning. The planning assumes that technology is static and I think that’s something we could do better.
And I guess a lot of the technology, the technologists, are at the moment primarily concerned with market share more than anything else.
Yes, and market share is the first question, but then the implication of the changing technology on behaviour and on how the system is used as an obvious follow-up. There are a number of predictions out there that autonomous vehicles will lead to increased suburbanisation, which is fine potentially, if people in the suburbs don’t have outsized impacts on the environment or whatever. But that does lead to increased travel and is contrary to the regional plans that we have in our cities, which call for a higher concentration into multi-family development.
So I think there’s very much a divergence between where the market is pushing and where the planners want to pull policy. And I’m not sure that that will be easily rectified.
We need everything going in the same direction. You’ve covered a fair bit of ground in your academic career. Is there an area of transport in which you haven’t done any work yet, that interests you and you’d like to do?
There are many things that are interesting, it’s just a question of time. And I’m a professor, I can work on whatever I want to work on really. So I don’t know that there’s anything that I’d want to work on that is more interesting to me than what I am working on, otherwise I would already be working on it.
I think there are a lot of interesting things going on in the Artificial Intelligence space, machine learning, in terms of, making predictions. I think there are interesting things that can be done with modelling. I don’t mean large-scale regional modelling, but just making predictions from big data. We’re doing some work in that area, and it is always exciting to be able to play with new datasets. These are datasets that 10 years ago didn’t exist and now they’re just streaming out all sorts of measures about GPS tracking, smart car datasets, all these kinds of things that we can access now, that just weren’t available before.
So work in that area would be cool. And it’s things I wish I had when I was 22 years old and starting out. Now the data that we have at hand is more than we imagined we could possibly have back in the day.
Okay, closing question. In the short term, next three to five years, what is it that’s heading our way that you’re most excited about?
The increased deployment of AVs and how will that work. And I’d probably put a five-year timeframe on that rather than a three-year timeframe on that. But we already see, for instance, in Chandler, Arizona, Waymo is operating driverless vehicles on relatively low-traffic, suburban streets.
But over time, they’ll get into more and more ambitious use cases and they and other companies are working on city streets in San Francisco and other places that are much more congested and where the driving environment is much more complicated. So I think there’s a lot interesting going on there and we’ll learn a lot about how the technology might work and where the weaknesses are and how we need to regulate the vehicles themselves and how we need to shape the environment to make those vehicles successful in an urban world, along with work on how do we make autonomous vehicles socially responsible.
Right now people are talking about connected vehicles, but connected vehicles have many weaknesses, latency, security risks and so on, associated with them and I think it’s a mistake to be thinking about connected and autonomous vehicles in the same breath, much less the same acronym.
But that’s not to say we can’t make autonomous vehicles that aren’t aware of the world around them, but now they act in a selfish way. Just as individuals route on roads, in a way that minimises their own travel time, autonomous vehicles will, for instance, in the absence of external requirements or incentives, choose lanes that will minimise their own travel time. But if we could get them to behave in a way that is more socially beneficial, that is, only change lanes if they have to or if it reduces congestion for instance, by moving from the more crowded lane to the less crowded lane, we might be able to do something that’s socially beneficial, but something that’s is selfless in a way that the current algorithms are not.
And the question is how do we incentivise the vehicles or the algorithms to do that, and maybe it has to be required or built into the vehicle’s operating system. Or maybe we could use a price signal, if there’s a monetary penalty to changing lanes, then people only change lanes when they have to rather than opportunistically changing lanes continuously. Because lane changing is one of the major causes of shockwaves and congestion and traffic and to the extent that we can reduce that, we can benefit from a traffic perspective.
We are doing some work in that area. And I think there are many opportunities to look not just at the lane-changing behaviour, but also the optimal gap from a car following perspective, when you’re following an autonomous vehicle versus when you’re following a non-autonomous vehicle. Mixed traffic is going to be with us for decades before we get an all-autonomous fleet, so how we handle driving in mixed traffic is still a wide-open question.
More from iMOVE Australia