Is a traffic tax the solution to congestion in cities? Not necessarily, according to research from Wharton real estate professor Gilles Duranton. He recently spoke to Knowledge at Wharton about the findings of his study, which is being funded by the Mack Institute for Innovation Management.
An edited transcript of the conversation follows.
Knowledge at Wharton: Could you start by telling us about your research?
Gilles Duranton: My research interest is in congestion because we all perceive that congestion, when we live in large cities, is a big problem in our life, right? It prevents us from going places when we want to. It is making our commutes longer than perhaps they need to be. In time use surveys, when we ask people what is the thing that they dislike the most in their life, they usually mention congestion as number one. Perhaps surprisingly, they mention childcare together with household chores as number two.
To assess the social cost of congestion, I looked at one particular city, Bogota, Colombia. I chose Bogota because I think we know far less about congestion in developing countries where it may be an even bigger problem than in developed nations. The way I think about the problem is the following: When you take your car and ride, you pay a time cost. One component of it is the time cost of going from one place to another without traffic. Then, there’s the traffic of others that slows you down. But you also slow down everybody else.
So there is a difference between how much you pay — what is the time you spend while traveling, which is your private cost of going places — relative to actually how much you cost to society by imposing more congestion on others. So, the others are imposing some congestion on you but you impose some congestion on others. [Since we think that more travelers cause travel costs to increase, the cost we impose on others is larger than the cost they impose on us.]
To quantify these costs, I start with data from an actual travel survey. I need to know where people go. On the supply side, I’m interested in how much slower traffic gets as you have more drivers. I’m trying to compute that quantity first. That allows me to infer both the private and the social cost of driving.
“The social cost of congestion is much smaller than we think.”
Of course, I am not looking at all the downsides from traffic. You also have really important issues such as pollution, accidents, and so on and so forth. But I’m only interested in congestion. That’s what I can measure, so that’s what I do. [This said, I only look at forms of transportation that take place on roads — cars, taxi, and small buses — but ignore large buses with rights of way since the congestion for that form of transit is different.]
On the demand side, I also need to know how much people pay for their travel in terms of time, when they choose to travel. But in order to know about that demand, I also need to know how much they would pay at times when they’re choosing not to travel. [In essence, I need to know at what price people buy transportation when they buy it and what the price is when they choose not to buy.] In order to get those counterfactuals, I’m scraping data from Google maps.
At the end of the day, what I find is that the demand for travel is mildly elastic. I think this is the first estimate in the literature, so hopefully that’s useful even though this first finding is not that surprising. But what is more surprising is what I find on the supply side. The wedge between the social cost of traveling and its private cost is actually far less than what we suspected, in the order of 5% to 20%, depending on the state of traffic. This is in contrast to the 100% or more that we were conjecturing before on the basis of studies of one particular road, or on the basis of some purely theoretical thinking.
Knowledge at Wharton: What were your key takeaways from this research?
Duranton: My key takeaway is that actually the social cost of congestion is much smaller than we think. I am not saying that traveling in large cities is easy. I’m just saying that the pure social cost of congestion is actually pretty small.
[With a wedge of 5% to 20% between what we pay in time and what we impose on society, an optimal situation would not be that different from the one we currently observe. An optimal congestion tax would reduce traffic, but not that much. There is huge demand for travel, especially at peak hours. Yes, we could make traffic extremely fluid at all times by drastically reducing the number of travelers but that would be socially unproductive given that demand would no longer be served.]
All that means is that for a city like Bogota, and maybe for other cities, we may want to think beyond just curbing congestion by imposing a congestion charge. This is still a meaningful idea but perhaps not the game changer we thought it could be.
At the same time, maybe the real issues behind slow traffic are elsewhere. We may want to be thinking about how much roadway should cities provide and what sorts of transportation mode choices we want to give to people, whether it should actually be private vehicles or more public transit, or a mixture of the two [and in which proportion].
We also want to pay more attention to whether and how we should manage traffic, in terms of coordinating red lights. So, some of the big issues seem to be about the nitty-gritty of traffic management, beyond trying to curb demand.
Knowledge at Wharton: Did any conclusion surprise you?
Duranton: The size of the social cost of congestion is much smaller than I expected. It’s also much smaller than everybody else expected, who are interested in transportation. As a result, I’m getting people who [are showing] slight disbelief.
At the same time, I think my findings are completely reasonable. The speed of traffic at the worst hours is only about half what it is at the best hours of the day despite the number of travelers being higher by a full order of magnitude. That suggests a wedge between private and social costs of perhaps 10%.
When you do, in technical terms, your “triangle” of welfare loss, you have a small wedge, multiplied by a small quantity divided by two, because that’s a triangle. It’s going to be at most 1% of the daily income that we actually lose in pure congestion.
So, what’s the problem? Again, there is strong demand, people want to travel and there’s only limited capacity for them to do so. That makes travel costly. But the pure external effect element is not that large.
Knowledge at Wharton: What are the practical implications of your research? What can other large cities do with this information?
Duranton: Actually, I’m revising, again, my judgment about how much of a priority should [be given to] charging for congestion. Like most economists, I thought, “OK, that’s what we need to do. That’s the beginning and the end of everything, about urban traffic.” I still think it’s an important idea and we should be thinking about that. But I now also think we should be thinking more seriously about how much roadway provision we have in developing cities like Bogota, actually, where the roadway may be grossly insufficient.
Imposing a congestion charge “is still a meaningful idea but perhaps not the game changer we thought it could be.”
We should also be thinking about traffic management very seriously, much more seriously than we’ve been doing so. [Urban planners, economists, and locally elected officials often make fun of traffic engineers who live to maximize throughput congestion at intersections. While there is a self-defeating element to this mission since better traffic begets new traffic, we should take this more seriously and join this up with other elements of urban transportation policy.]
Knowledge at Wharton: What sets your research apart from prior work in this area?
Duranton: There’s really not much prior work on congestion that actually looks at congestion rather than speculate about it. You are going to tell me “You must be kidding.” What people have done previously was measure things at the level of one particular road, or a set of segments on a road, or a small number of roads. It is deeply problematic to do that because on a road, when you keep adding cars, at some point traffic will stop, right?
In an entire city, it’s a very different proposition because what happens somewhere has an effect on what happens elsewhere. It’s a network — things spill over — and we need to think about this network in its entirety. In particular, when the highway is clogged, people will start using main arterials. When the main arterials are clogged, people will start using secondary arterials, etc. It means … there are nearly always more options for going places.
This essentially puts a floor [on] travel speed. Local roads are going to be slower; they’re not going to be practical. That may have some negative implications in neighborhoods, but people will start using them. It means barely ever will traffic come to a complete standstill in an entire area. Again, it can happen when there’s an accident on a particular road. But it normally does not happen.
Knowledge at Wharton: How will you follow up this research?
Duranton: What we did here was to work on only one city, where we had a lot of precise information. What I want to do now is actually work on all major cities of the world. We started collecting data for about 150 large cities in India and we want to generalize that to 4,000 to 5,000 cities all over the world so that we can actually look at travel behavior and the ease of travel, and the ease of reaching some particular destinations such as schools, train stations, or hospitals in cities of the world.
Then, the objective will be to assess the determinants of these measures of mobility and accessibility such as the level of advancement of the country, the amount of roadway, particular policies, and all these things.
Knowledge at Wharton: How do you foresee innovations in vehicle technologies reshaping our cities based on your research?
Duranton: The first point is that self-driving cars, or whatever you may want to call them, they are coming, and they’re coming reasonably soon. The first big issue is, at some point we’ll need to get rid of the steering wheel, which is a big discrete change relative to what our cars are currently. This change, which is technically possible, may not go all that smoothly.
[Google is betting on that radical change, but car producers want something gradual. There will be a huge regulatory fight at some point in the future.] But I think within 10, 20 years, that will happen. So the first thing that self-driving cars will do will be to lower the time cost of travel. Not because it will go that much faster. Actually, they will probably be slower. But we can do lots of other things while traveling instead of driving.
“Self-driving cars will … lower the time cost of travel. … We can do lots of other things while traveling instead of driving.”
Actually, when we try to estimate the costs of travel, a large fraction of that cost, with 50% to 90% depending on your income, will be your time. Now this time can be used productively, instead of just driving and paying attention to traffic. You can actually do work, watch movies, and do all sorts of things. It means that actually, the time cost of driving will be far less. So when we look at what is the propensity of people to travel depending on how costly it is to travel, we can only guess that people will want to travel more — and perhaps a lot more. Maybe 50% more, maybe 100% more.
We’re talking possibly about a lot more driving, which means that people will probably be willing to come to work from much further away. I do suspect that indeed, there will be a big movement of suburbanization that will go even much further than anything we’ve seen. Some people are deeply worried about that. That’s clearly a distinct possibility with this new technology.
This new technology will reduce the cost per mile. What you’re also going to reduce is the necessity to park, to go and reach your car, and so on and so forth — all what we can call the fixed cost of driving. In that case, that will mostly benefit people that actually pay high fixed costs per trip. These are people who live in dense urban environments, where finding parking is difficult, where accessing your car is difficult. These people also will gain a lot. They will start sharing cars.
Indeed, cars will start showing up exactly where you are more or less when you want them to show up. One estimate says that we spend about 30% of our time when driving in the center of cities looking for parking. This will disappear. So that will also be a big gain. It means that city center living will also become more attractive.
It’s unclear which one will gain most. But I think we will see both at the same time: more suburban development really far away, but also greater attractiveness of city centers at the same time.
Knowledge at Wharton: How do you think self-driving technology will impact congestion in major urban centers?
Duranton: I think self-driving cars will actually impact major urban centers in lots of ways. What’s going to happen is that self-driving cars will be much more agile at going through intersections. Road capacity will increase a lot — the distance between any two cars, which these days is about 1.5 to two seconds — can be dramatically reduced. Self-driving cars can travel really close to each other. This means that road capacity will be tremendously increased, perhaps by a factor of four or five, and on highways, maybe 10.
It means that we may actually need, in the end, less roadway, even though I think traffic will increase a lot, but not in those proportions. It means that congestion may become a thing of the past at some point, which would be great, of course. We may also have very different cities in terms of layout.
If everybody has a self-driving car — and that will happen at some point. Maybe not 15 years from now, but 30 or 40 years from now — we will no longer need traffic signals and so on and so forth. We may not even need sidewalks. We may actually be in for much more pleasant urban environments.
