Study: More Roads = More Traffic

Rigorous academic work finds that neither road construction nor new transit can solve congestion.

Today’s news carries a story that I’ve been expecting for a while: “Connecting Washington,” a task force convened by Washington’s governor, has called for $21 billion in new transportation investments over the next 10 years. I haven’t seen the recommendations themselves, only the news report. But it looks like the money would get spread around a bit—with some for ferries and some for transit—but from what I can gather, most of the money would be slated for roads.

So in the upcoming months, I expect we’ll be hearing a lot about how investing in new roads will help clear up traffic problems, particularly in greater Seattle and the Washington side of greater Portland.

But a study that’s been sitting on my desktop for a while—which I count as the single most interesting transportation paper I’ve run across all year—suggests otherwise. Co-authored by researchers Gilles Duranton and Matthew A. Turner from the University of Toronto, it’s a careful and remarkably thorough analysis of the relationship between urban highway space and traffic volumes in the US. And its key finding is straightforward:

For interstate highways in the densest parts of metropolitan areas we find that vkt [Vehicle Kilometers Traveled] increases in exact proportion to highways

In short, the authors find that building new urban highways simply increases traffic volumes—not in some general, intuitive sense, but in the sense that, all else being equal, a one percent increase in urban highway space increases urban road travel by precisely one percent.

To many folks who follow transportation issues closely, the idea that new roads create new traffic is familiar—yet it sometimes gets bandied about rather casually, as if it’s something that we know but don’t particularly need to prove. But there’s nothing casual about Duranton and Turner’s analysis. It’s thorough, well thought out, rigorous, employs careful controls to separate cause from effect, and uses the best available data for the US transportation system from (I kid you not) 1835 to the present.

And while road skeptics will undoubtedly be heartened by the findings about highways and traffic, they may be troubled by another of the study’s findings: public transit has virtually no effect on traffic volumes.

The irrelevance of transit service to urban traffic volumes goes against both intuitions and the claims of many transit analysts. Yes, it’s easy enough to argue that if all transit riders started driving, then they’d clog the roads during rush hour; and conversely, that if a bunch of people switched from cars to buses, rush hour traffic would get lighter for a while. But as the authors point out, if drivers switch from cars to buses or trains, it has much the same effect as adding new road space: traffic clears up temporarily, but faster travel quickly attracts more drivers who take longer trips.

(And just to be clear—and before the haters start hatin’—the authors aren’t arguing that transit is irrelevant to transportation, or that it has no value to society. I certainly believe that transit has substantial value for people who can’t afford a car, can’t drive, prefer not to spend their time behind the wheel, or want to live in a place where road space is constrained. Saying that transit doesn’t have much systemic impact on congestion isn’t the same as saying that transit isn’t a good idea!)

So if this study is accurate, then it poses some serious problems for transportation compromises that couple road expansions with transit projects. It’s easy to convince yourself that those kinds of compromises are win-wins—that road expansions will simply ease congestion, and that transit will help keep traffic growth in check. But this research suggests that that sort of compromise is a pipe dream. What matters to highway VMT, according to Duranton and Turner, is how much free road space we create. Just about everything else is a distraction.

I’ll leave the technical details to the avid reader (you really should read it!) and simply pull out two quotes that explain some of the the other implications of the research.

We find that the welfare gains for drivers of building more highways are well below the costs of building these highways. This conclusion follows, not from the high elasticity of vkt to roads, but from the fact that new roads do not reduce the cost of travel sufficiently.

Road projects are sometimes sold to the public based on promises that easing traffic congestion will save everyone time. But the study finds that these congestion reductions are short lived at best. The additional people on the road get some value from their trips (otherwise they wouldn’t make them!) but for society as a whole, they find that the cost of new roads tends to outweigh the benefits.

We…identify three important sources for this extra vkt: an increase in driving by current residents; an increase in transportation intensive production activity; and an inflow of new residents.

Interestingly, they find that boosting driving among existing road users is only one way that new roads can induce new traffic. Road space also changes the mix of businesses in a metro area, with more road space leading to more driving-intensive businesses; and they even find that lower commuting costs can attract new residents. The latter effect is small—they find that a 10 percent increase in road space can lead to a 1.3 percent increase in population over a decade, all else being equal—but still significant.

I certainly don’t mean to hold up this study as definitive. It’s only one piece of evidence among many on the effects of transit and road construction on traffic. Yet it’s a particularly well-researched piece—one that folks looking to understand transportation (rather than simply argue about it) would be unwise to ignore.

We are a community-supported resource and we can’t do this work without you!



  1. Morgan says:

    Thanks Clark.
    I’m wondering how this can fit into our advocacy for healthy transportation systems. One the one hand, we’ve known this sort of thing for years, even if not to such a fine degree as this study makes clear. Yet, it doesn’t penetrate into general conciousness. Why?

    Is it because the latency from capacity expansion to the return of congestion takes to long for us to perceive it? It is because we don’t want to know, and we’d much prefer to maintain our habits to the point of dissonance with our values and reality?

    What is going to crack this nut in the public sphere?

    Again, thanks for these many months of exploration into traffic demand.

    • Clark Williams-Derry says:

      Gosh, I don’t really know, Morgan, but I think you’ve got a good part of the explanation: we’re a short-sighted species, and poorly adapted to think about the behavior of complex systems.

      Besides, when evidence conflicts with our opinions, it’s fairly typical for the facts to bounce off while the opinions stand firm. As J.K. Galbraith once said, “Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof.”

      • Morgan says:

        love the quote

  2. Morgan says:

    Maybe, we’ve just been trying to solve the wrong problem–congestion.

  3. Stephen Rees says:

    It is not news. It is not just one study. Over the years there have been many of them. The obverse is also true. Removing road capacity does not make congestion worse.
    (Cairns, Hass-Klau and Goodwin 1998 – a meta study of 48 road capacity reductions around the world)
    Traffic expands – and contracts – to fill the space available. All our models still reflect the false notion that traffic is like water, that must be accommodated. In reality it is more like a gas than a fluid. Building new roads – or making more road space available – induces more trips. Yet no model has an algorithm that shows that.

    • Clark Williams-Derry says:

      I love the gas-vs.-liquid analogy!

      And while I agree that there have been lots of studies and meta-studies on induced traffic over the years, this one struck me as particularly well done and relevant to today’s debates. Plus, by taking a broader and longer-term look at the data, they found an elasticity of traffic to road space of 1, with causal linkages. Many previous studies, often specific to particular roads or corridors, had often found that elasticities were positive but <1.

  4. Jane says:

    The traffic situation in many big cities is becoming a big problem. Traffic jams are annoying but what really drives me crazy is a lack and high cost of parking lots in Vancouver where I live.

  5. Jane says:

    This isn’t really new news, but a confirmation of what we’ve sensed for some time now. Road building should be understood as a job-creation program, which it is because of its labor-intensity and large “multiplier effect” in the local economy. That road-building is sold as congestion-relieving serves merely to assure the public that more folks will benefit from it than only the construction crews who receive the jobs. Spending on public transit isn’t as effective as road building in creating jobs and retaining the multiplier effect within the local economy, thus making it a harder sell.

    • Clark Williams-Derry says:

      Huh. I’m wondering where you get the data showing that roads have a higher jobs multiplier than transit. This study from UMass Amherst shows the opposite.
      Similarly, this study found that road-only infrastructure has the lowest jobs multiplier among the transportation projects they considered (though the comparison was bike-ped infrastructure, not transit.)

      • Morgan says:

        Clark, did I miss something? The first UMass study doesn’t seem to address the employment effects from roads at all. Rather, it compares many project types to coal & petro production.

        On the second study, the employment effects from building infrastructure are address, but I didn’t catch a mention of project utility as, say, an input to something else, like economic activity or social endeavor. This is an area where we stumble a little, so I’m wondering if some have studied these elements.

        For example, it’s harder to argue that bike lanes have an impact on economic productivity than might an additional freight spur to the port. Maybe, I’m thinking in the wrong direction.

      • Clark Williams-Derry says:

        Drat, Morgan, I think I may have given the wrong links. I thought the road comparisons were in there.

        A couple of recent studies (see here and here present numbers on roads vs. transit job creation. Maybe I was thinking about those. But I thought I recalled seeing something from PERI.

  6. John Newcomb says:

    A hundred years ago, Victoria’s inter-urban tram system was key to the birth and development of exurbs far from the city-core – and then came the auto to speed-up the process of suburbanization.

    Perhaps one direction that Duranton-Turner research might point to is do not seek to reduce congestion by either more road space or transit, but to re-jig the urban form itself. Like more complete satellite cities.

    Of course, the political will to challenge the current urban form would need to be far greater than just voting fund increases for road building or more buses.

    • Morgan says:

      I tend to agree about the lack of value in reducing traffic congestion with the supply side solutions. I think the key is to reduce demand, like reducing the distance between destinations, eliminating trips, and many of those cool commute-trip reduction programs. It’s kinda like, do we want to learn how to feed more people or learn how to have fewer people?

  7. David says:

    And so… Let’s invest most in the form that pollutes the least. That seems like the no-brainer.

  8. John Niles says:

    It’s great that Clark is shining a light on this work by Duranton and Turner. It got me looking up more by these authors!

    The working paper dated 2009 that Clark has found achieved publication just this past October in the peer-reviewed American Economic Review The published article has stimulated quite a bit of recent blogging on what it says, such as and Clark is making an interesting run of his own through the arguments in his essay above.

    These blog discussions are easier to read than the mathematics in the Duranton and Turner paper. Perhaps the easiest next step of all if you want to dig in with what the Toronto academics are saying is to read a commentary essay on the topic by Duranton himself for a prominent, non-partisan DC think tank, Resources for the Future:

    It’s been known for at least twenty years by researchers and some commentators that road expansion, transit expansion, and telecommuting expansion don’t reduce traffic congestion and vehicle miles traveled.

    These three sources of hope are all related to the point that in busy urban environments, empty and free road space — and free parking lots near busy places — gets filled with cars as time goes on.

    At first there’s lots of space, like I-90 across Mercer Island for a while after it opened. But as time goes on, land development activity along the I-90 provided more and more reason to use that highway and it became crowded.

    I wrote a summary on the telecommuting contribution to taking cars off the road in the early 1990s in a Department of Energy report Beyond Telecommuting, . I included reference to a book that got the discussion started for me and many others, Stuck in Traffic by Anthony Downs. He has been a speaker at leadership conferences in Seattle in earlier years, and always talked about the difficulty of reducing congestion. He suggested putting a microwave oven in your car so at least you’d have hot food while stuck in traffic.

    Sound Transit and its fans suggest sometimes that the billions of dollars in railroad infrastructure ST is building will reduce traffic congestion, but the environmental studies they put out say otherwise. In only a few cases have Sound Transit agency staff/consultants implied directly on paper that congestion will go down if it builds the rail lines it plans to build, and it’s generally the result of convenient errors in analysis that are stated too obscurely to be successfully challenged.

    What Sound Transit is offering is a way to avoid road congestion — take the train. Of course as anybody who has stood on a crowded transit vehicle or station platform knows, sometimes you are trading road congestion for shoulder-to-shoulder congestion with your fellow commuters up close and personal.

    When King County Metro and its friends in politics named the new $20 car registration tax last summer a “congestion reduction charge” they could claim that they meant to make the aisles of the buses less congested from too many standing patrons. Nothing noticeable will change on the roads.

    The potential of 360 degree sensors in cars and car-to-car wireless data communications to expand road lane capacity and reduce congestion-causing accidents is likely to change the dynamics of road utilization quite a bit, I and others would predict. I don’t mean people talking and texting on cells phones, although that will be safe when cars steer and stop by themselves. By 2030.

    Also, congestion sensitive road use fees — variable by the mile fees on every car that make the roads more like a water or electric utility where you pay directly for usage — will be a game changer, a point supported by Duranton and Turner. As suggested on Monday by Transportation Commissioner Dick Ford at the Monday meeting Clark refers to at the top of his essay, these road use fees can be phased in gradually and voluntarily by motorists one at a time who accept a better deal than they get with gas taxes.

  9. Curtis says:

    I’ve always thought the promises of congestion relief from road or transit projects as crass manipulation by the proponents of these projects. Thanks to Clark, the studies to back this up are now available to arm-chair transportation enthusiasts like me. Hopefully, bringing this kind of information into the public discourse will also make it harder for transportation project proponents to tout their congestion reducing benefits and focus on some real issues.

    I fall in with John above that we should be looking at a future transportation system that is energy and time efficient. This implies planning that transcends the boundaries we currently have in place between jurisdictions and government programs responsible for roads, bus, rail, walk, bike, etc.

  10. cold, harsh reality says:

    Hey Clark,

    Data from around here show that supposed correlation is false. The East Link FEIS (page 7 of chapter 3) says this:

    “On the floating bridges, the average daily traffic volume is 140,000 to 150,000 vehicles. This consists of about 135,000 vehicles per day in the eastbound and westbound mainline lanes and about 15,000 daily vehicles in the reversible center roadway (WSDOT, 2007).”

    Since then traffic volumes have decreased significantly, a phenomenon that is contrary to what that report out of Toronto asserts should happen.

    The traffic volume on the I-90 floating bridges now is about 15% less than it was in 2007. The 2010 WSDOT daily traffic report shows that the average daily traffic volume is approximately 137,500 vehicles. This consists of about 126,000 vehicles per day in the eastbound and westbound mainline lanes and about 11,500 daily vehicles in the reversible center roadway. Those data come from page displaying the I-90 west highrise traffic volumes:

    Traffic is decreasing, which means extra highway capacity does NOT lead to increased vehicle use of that capacity. That whole “induced congestion” canard is false.

    • Clark Williams-Derry says:

      Hm. I agree that traffic is decreasing, or at least holding flat, on major roads throughout the Northwest. (That’s what our blog series, Dude, Where are My Cars? is all about.) Those trends do provide some evidence that there isn’t necessarily a one-to-one relationship between road space and traffic volumes. I’d describe I-90 traffic as roughly flat, based on my read of the state’s Annual Traffic Report series. Really, it was ~flat from 1998-2007, then fell in 2008-2009, and started rising again in 2010. We’ll see what happens in 2011 — maybe it’ll decline again, with higher gas prices.

      I hadn’t looked closely at the source you sent –but my preliminary look shows ~131K cars in the mainline lanes on I-90 at Shorewood Drive. But I’ll have to look carefully to figure out how that compares with previous years. (In my experience, the different traffic volume data series never quite seem to match up with one another. Even counts from adjacent counters seem to disagree…)

      The larger point is that congestion costs, toll costs, gas costs, demographic shifts, mpg trends, and other factors all interact to influence VMT trends. And when there’s lots of free road space, congestion costs go down: all else being equal, lower congestion costs tend to increase VMT. But right now, it’s just not true that all else is equal: we’re seeing stagnation or declines in real median incomes, declines in incomes for low-wage workers, rising gas costs, high unemployment, and some demographic shifts that are all depressing driving a bit, despite slight declines in congestion costs.

      So if you’re saying that road space doesn’t by itself fully determine VMT – I agree. If you’re saying that there’s no relationship between road space and VMT, I disagree.


      • cold, harsh reality says:

        I’m not saying either of those things you try to attribute to me, Clark. I didn’t even mention vmt, so it’s bizarre you would try to characterize what I posted as arguments relating to vmt.

        One point you should take away is the “key finding” of that report you call “straightforward” (that for interstate highways in the densest parts of metropolitan areas we find that vkt [Vehicle Kilometers Traveled] increases in exact proportion to highways) is demonstrably false.

        If it were true, there never would be a decrease in traffic, and traffic on the I-90 bridges has decreased about 15% since 2007. That’s undoubtedly due in large part to how there has been a significant loss of jobs in downtown Seattle.

        My main point is the “induced demand” argument is bogus . . . building highways does NOT inexorably lead to traffic filling up new highway capacity.

      • Clark Williams-Derry says:

        Sorry, I didn’t mean to attribute anything in particular to you. I simply didn’t (and don’t) understand what you’re saying. (Just to be clear: traffic volumes on a fixed segment of road can be converted directly to VMT on that segment, and vice versa; you just multiply or divide by the length of the segment. So for many purposes, statements about traffic volumes and VMT are essentially interchangeable.)

        Looking at the WSDOT data, I too see a decrease in traffic volumes on I-90 in recent years. But it’s much smaller than 15%. When I look at successive editions of the Ramps and Roadways report, one of which you link to, and match up specific segments with one another, the drop is about 5-6% depending on when you start your measurement. As an example, here’s a traffic volume series from the “east highrise” mainline segment of I-90, from the Ramps and Roadways reports.

        1996 133,300
        1998 135,590
        2000 136,655
        2002 131,314
        2004 133,570
        2006 135,750
        2008 131,530
        2010 128,970

        Perhaps the 140-150K number quoted by WSDOT is the weekday rather than the all-week total. (As I say, it’s often tough to line up all of the different statements about traffic volumes. “Per day” sometimes means Monday through Friday, and sometimes Sunday through Saturday.)

        More generally — I believe that “induced demand” is exceptionally well documented in the academic literature. But there is a very legitimate debate over the *size* of the induced demand effect: different researchers come up with different answers. The induced demand debate is just a subset of a larger literature on transportation elasticities — e.g., how fuel costs, tolls, time costs, income effects, etc. affect transportation habits.

        But I don’t expect I’ll be able to convince you that induced demand is real, just as I don’t think that you’ll convince me that it’s not. So perhaps I’ll save both of us some time, and just wish you happy holidays!

  11. Frank says:

    Not to get technical..
    but you can’t take specific roads, or bridges, etc. and deal simply with them.
    All sorts of external factors click in when talking about, for example, the I-90 floating bridge:
    General economy on both sides of the bridge; ie jobs availability, closure of shopping centres, construction of new homes.
    Movement of people in and out of the areas involved.
    I think the general premise of the author and article remains intact.

    • Clark Williams-Derry says:

      I absolutely agree, Frank!! The elasticity figures found by the authors are derived from large sets of data — but I’m certain that there are lots of local peculiarities, with particular road expansions that don’t correspond to traffic increases, and some expanded roads where traffic grows even more than expected.


Leave a Comment

Please keep it civil and constructive. Our editors reserve the right to monitor inappropriate comments and personal attacks.


You may add a link with HTML: <a href="URL">text to display</a>