As urban areas become more congested, and concerns over the environmental damage that can be caused by vehicle emissions grow, many municipalities are adopting land use guidelines that encourage compact development. A city with greater density, they theorize, will reduce the need to drive by bringing services and retail closer to the areas in which people live.   

But new research co-authored by Wharton real estate professor Gilles Duranton finds that such policies may not have as great an effect as planners believe. In “Urban Form and Driving: Evidence from U.S. Cities,” Duranton and Brown University professor Matthew A. Turner find that increases in density cause only minimal decreases in aggregate driving, meaning it is unlikely to be a cost-effective policy for responding to traffic congestion or automobile-related pollution.

Duranton recently talked with Knowledge at Wharton about their findings and how they might be applied by businesses or governments.

An edited transcript of the conversation appears below.

Urban Form and Traffic

I’m interested in understanding how what we call “urban form” affects traffic behavior. By urban form, I mean all sorts of characteristics of the built environment that you live in — densities, the amount of jobs relative to residents, the quantity of roads and all of those characteristics. We’re interested in how this is going to affect the way you travel and how many miles you actually travel, especially by car. Why are we concerned by that? Because there are some social costs associated with using personal transportation. We know for each mile that we drive, society incurs a cost of three or four cents — maybe slightly more — associated with two different forms of pollution. One is greenhouse gas emissions — i.e., carbon that fosters climate change, global warming and all of that. And the second one is much more localized: small particulates, which could affect people’s health.

Key Takeaways

We found that urban density, measured by the density of residents and jobs within a 10 kilometer radius of where you live, appears to have small causal effect on driving. So, if you bring up density by about 10%, it leads to reduction in traveling of about 1%.

Surprising Conclusions

What surprised us most is the fact that we were expecting the results to be fairly complicated and fairly subtle. What we discovered is that reality was, for once, very simple in the sense that there’s one major characteristic of cities that matters: the density around you. Everything else is really secondary and matters far less. And what we also found is that those results were extremely robust.

“For each mile that we drive, society incurs a cost of three or four cents — maybe slightly more — associated with two different forms of pollution.”

For instance, we expected different groups of residents to react differently depending on where they lived — for example, we expected people who like driving to live in places where it’s easier to drive. Actually, these differences were not at all important in our results. Everybody seems to be reacting more or less in the same way. Or at least, we could not find very much of a difference across different groups, across different types of environments. So basically, again, if you bring up density by about 10%, you reduce driving by about 1%. 

‘Everything Else Will Not Do Much’

Even though we do find some effects, with 10% higher density leading to 1% less driving, it’s actually a very small effect. If you want to understand what that means in reality, [imagine] you were to take the 30 million Americans that live at the lowest density in rural areas and bring them to the highest density areas in the country. That’s a displacement where the scale is not possible at the moment — it’s massive. It’s not something that’s going to happen at all. But that would lead to a reduction in driving by 4% or 5%. And something that is still drastic, but slightly less so, would not have that much impact: If we took, again, those 30 million Americans from low-density areas to high-density areas, but left some people in those low-density areas, the effect on driving would be minimal. Those that are brought to high-density areas will drive less. But those that are left in the low-density areas will drive lots more, and it’s nearly a perfect offset.

So, as a result, you won’t be able to achieve very much in terms of global warming using urban policies of that sort. To go after local pollution, you need a tax for congestion — i.e., the concentration of traffic in some areas of a city — so you need to make drivers pay for that. And you need to tax carbon emissions. For instance, the province of British Columbia does this in Canada — it’s a resounding success. I know it’s particularly a really hard proposition in this country, but that’s the one thing that works. Everything else will not do very much.

“I know [taxes are] particularly a really hard proposition in this country, but that’s the one thing that works. Everything else will not do very much.”

Global Problems, Global Solutions

We have this perception that the federal government in this country is not very effective at doing very many things and that a lot of the solutions to large-scale problems have to be local…. But global problems require global solutions; they require federal interventions. We are not going to solve global warming by having cities practice small growth. That will only achieve very, very little. And some of it, like imposing “green belts,” if anything, might be just counterproductive.

What Sets the Research Apart

This is an area where there has been a lot of research. What is setting us apart is first, we use extensive data for travel. Instead of looking at a tiny area and the number of cars, we actually rely on a big survey done by the Department of Transportation with nearly a million trips. That’s a very large-scale survey. Then, we use extensive measures of the local environment where people reside. We use satellite imagery to see on a one kilometer by one kilometer basis the [details] of the physical environment, and also use a vast array of data coming from the Census, coming from employment and so on and so forth. We combine all that to know exactly what there is around where people live, in terms of number of jobs, the number of residents, how rich those residents are, how educated they are, and so forth. So, we use vastly more data to describe the environment.

Also, just because you observe a negative correlation between how dense an area is and how little people drive, [you can’t assume] that this relationship is causal. This relationship may just be completely spurious, because again, people who like to drive may want to live in low-density environments. So what we’ve done was to design a new methodology to address that question…. We look at people, how they responded to differences in the amount of population in an area, for those that have stayed for a long time. That allows us to go through the selection problem and actually make some causal statements where we can actually claim that the local environment has an effect on the amount of driving that you do, but this effect, again, is small.

“Global problems require global solutions; they require federal interventions. We are not going to solve global warming by having cities practicing small growth.”

What’s Next

I want to think more about congestion, because we think of congestion as a big problem. It’s true that traffic jams are there for everyone to see; they’re everywhere, particularly in developing countries where people keep buying cars with economic growth, and the situation gets worse. At the same time, we have no serious idea about the economic cost of that congestion. And for that, I want to be able to use new data that’s now made available through mapping services, like Google Maps, where you can actually get some travel times from anywhere to anywhere else in real time, with information about the number of travelers. So, you can actually estimate the true cost of congestion.