The Way We Use Traffic Projections is a Lie. Even at the National Academy of Sciences.

The Infrastructure Cult Comes to the National Academy of Sciences

At Strong Towns, we're no strangers to flawed "studies" that attempt to drum up support for massive increases in highway and other federal infrastructure spending. Many of these studies are essentially propaganda for a group of lobbying organizations that directly benefit from such spending. We've called that clique—including the American Society of Civil Engineers (ASCE), Inrix and the Texas Transportation Institute (TTI), among others—the Infrastructure Cult.

But a colossal new report, commissioned by Congress in 2015, on the future of the Interstate Highway System is troubling because it doesn't come from one of those advocacy groups with an obvious agenda, but from the Transportation Research Board (TRB), a program of the National Academy of Sciences. This is an institution that you might expect to be committed to high standards of empirical rigor and disinterested truth-seeking. They were tasked with producing a far-reaching report based on a broad range of expertise. And yet they've produced a document, called Renewing the National Commitment to the Interstate Highway System: a Foundation for the Future, that ends up mostly reproducing the conclusions of Infrastructure Cult propaganda.

The study—all 615 pages—includes a colossal amount of detail on nearly every aspect of the interstate system, from pavement condition and wear, to safety, to the effect of interventions like tolling on traffic levels, to the way shifts in regional population (migration to the South and West) have affected demand for freeway travel. No stone is left unturned, and much of the content is apolitical and uncontroversial.

And yet, the whole thing is set up in a way that reflects the biases of the Infrastructure Cult, and makes it a foregone conclusion that the interstate highway system must be not only maintained but significantly expanded. And sure enough: The study concludes that we should more than double federal interstate spending to as much as $70 billion a year, with $13–$31 billion of that devoted to new construction (not maintenance or replacement).

On the contrary: there is a compelling list of reasons we should think twice before adding a single new mile to a single new lane of urban or suburban freeway, anywhere in the U.S., and this report from the TRB actually directly acknowledges most of those reasons. None of them end up mattering, though, in the face of the secret weapon of every Infrastructure Cultist: the all-powerful traffic projection.

The Problem With Traffic Projections: Chicken, Meet Egg

The study's mandate from Congress was to produce "recommendations on the actions necessary to restore and upgrade the system to meet the growing and shifting demands of the 21st century, through the next 50 years." The primary tool through which it measures those "growing and shifting demands" is traffic projections.

Figure 1: The TRB study models traffic as a one-directional result of economic and demographic trends, but does not adequately account for how road expansion itself induces more driving.

Every highway expansion project has a traffic projection behind it: a forecast that says, "x number of people will be using this road 20 (or 10, or 15, etc.) years from now, and its current capacity can't handle that future traffic." Ergo, the road needs to be expanded, or alternative routes do.

The problem is when it’s the expansion itself that turns those traffic projections into a self-fulfilling prophecy.

Traffic growth is a two-way feedback loop: building more roads actually causes people to do more driving. Intelligent people who study transportation networks all understand this. It’s called induced demand, and it’s not a controversial or unorthodox idea. A famous paper from 2009 by Gilles Duranton and Matthew Turner actually concludes that not only do new urban freeways induce new traffic, they do so in a near-perfect 1-to-1 ratio.

Let that sink in. Traffic growth on freeways is almost perfectly proportional to the expansion of the freeways themselves. The reason: transportation infrastructure, like adding lanes to a freeway, affects land use. It affects where people choose to live or locate their businesses, and where developers choose to build. That land use, in turn, affects traffic, through millions of decentralized decisions that all inform each other as well. The end result is usually that the new road capacity just allows people to drive more miles to reach the same stuff—jobs, schools, shopping, etc.

The TRB study includes data that makes this driving-more-miles effect clear. Here’s a chart from the study of vehicle miles traveled over time in the U.S.:


While rural interstate travel has ticked up more modestly since 1980, urban freeway travel and urban non-freeway travel have both more than doubled in a generation. According to the TRB study, from 1945 to 2005, VMT (vehicle miles traveled) increased at 4% per year while population grew at only slightly more than 1% per year (and GDP grew at 3%). It’s evident from the large body of research on induced demand that suburban expansion accelerated by freeway building is a huge cause of this explosion in American driving. But the TRB study downplays the induced-demand effect, and treats traffic not as a complex feedback loop but as a one-way mechanism: regional economic and population factors cause traffic to increase, and the only role for roads is to respond to the increase with more capacity.

The authors don’t do this because they actually believe it’s that simple. They literally state that they do it because it’s too hard to do it the other way:

As discussed in the previous chapters, highway planners have had mixed experience forecasting VMT. Planning of investments in long-lived transportation infrastructure, however, requires VMT forecasting.

The committee concluded that existing [Travel Demand Forecasting] models do not offer the national- or regional-level prediction capabilities needed to assess system- level impacts from Interstate investments.... Because there are no existing tools with which to analyze these demand responses at the transportation network level for the entire country or regionally, the only alternative was to consult the recent history of travel behavior as indicated by past VMT growth rates to develop a reasonable range of future VMT growth rates to apply to the… models."

Translation: "It's impossible to really predict how traffic will change, so we're just going to assume that because it's been rising, it's going to keep rising."

Perhaps that's uncharitable: they do put some work into estimating the rate at which traffic will rise, and provide lower- and upper-bound estimates. (The lower bound, they state, reflects what might happen if we more aggressively implement measures like congestion pricing and toll lanes to curb demand for existing roads.) But even their most conservative estimate has the system growing. Any policy choice that would aim to halt that growth or even reduce miles driven is immediately off the table as soon as you assume that.

Instead, by extrapolating mileage trends into the future, we are extrapolating the policies that caused those trends—like massive highway building—into the future as well, and we're subtly ruling out the option of changing those policies. We’re saying, “New freeways will be necessary to accommodate all the traffic generated by our new freeways.” Chicken, meet egg.

Joe Cortright at City Observatory has written an excellent take-down of the TRB report (one in a long line of excellent take-downs), and in it he shares my favorite cautionary tale of what happens when you assume traffic will grow forever because it has grown in the past: the saga of the Washington State DOT and the unchanging projection. There is never a bad time to share this graph, so here it is:


This is the insanity that results when you treat traffic like a law of nature and not the result of our infrastructure decisions.

Oh, and did you think this spending proposal would allow us to build our way out of congestion? Think again, says the TRB:

If the nominal VMT growth rate of 1.5 percent per year is assumed, investments over the coming 20 years will need to average $57 billion annually—more than double the spending level of today.

Even with these large expenditures, it is notable that some aspects of Interstate condition and performance would not improve, and perhaps deteriorate, relative to current conditions after 20 years. For example, if VMT grows 1.5 percent per year, even an average annual expenditure of $57 billion for pavement, bridge, and capacity-related improvements would result in levels of traffic delay comparable to today. To achieve much lower levels of delay (e.g., 75 percent of current peak-period hours of delay), as shown in Figure 5-7, would also require much higher levels of spending. Those much higher spending levels would not be justified on a benefit-cost basis, as calculated by the models.

Performance won’t improve, but let’s build the thing anyway!

By extrapolating mileage trends into the future, we are extrapolating the policies that caused those trends—like massive highway building—into the future as well, and we’re subtly ruling out the option of changing those policies. We’re saying, “New freeways will be necessary to accommodate all the traffic generated by our new freeways.” Chicken, meet egg.

Steamrolled by This Report: All the Reasons Not to Let VMT Keep Growing

As is typical of a mammoth document written with the contributions of hundreds of people, this report is not tightly focused, and interesting digressions have a way of sneaking in. For example, there are a couple overt discussions of the immense economic and social damage caused to urban communities (most of them poor and largely non-white) by urban freeway construction. But then the study undercuts this acknowledgment with the milquetoast conclusion, as Cortright points out, that the effects of freeways on the inner cities were “not entirely beneficial.”

Not entirely beneficial. Tell that to the residents of a neighborhood that was bulldozed for a freeway, or whose property values crashed after freeway-fueled suburbanization pulled the rug out from under the urban housing market. The damage to America's inner cities from interstate highway construction wasn't "not entirely beneficial." It was incalculable.

The damage to America’s inner cities from interstate highway construction wasn’t “not entirely beneficial.” It was incalculable.

Unfortunately, "incalculable" has a literal meaning too: not able to be calculated. This is because it involves too many messy, complex questions of human social behavior, feedback loops, and factors like systemic racism that helped spur suburban flight (for which the interstates were merely a necessary conduit). The response of engineers and policy experts who rely on data to this incalculability is to wash their hands of it. Just do a nice simple traffic projection instead.

A similar thing happens with regard to environmental costs. This study even devotes not one or two but several sections to climate change. You might think that if "climate change" and "interstate highway system" appear in close succession, it would be because the transportation sector is the largest source of carbon emissions in the U.S., accounting for 29% of them. Bafflingly, though, this study brings up climate for a very different reason: only to discuss the impact that a more erratic climate, with worse storms and heat waves, might have on roadway conditions. 

There's that simple, one-directional thinking again: the climate might affect how the highway system functions; land use might affect how the highway system functions; traffic growth might affect how the highway system functions; but nowhere are the authors allowed to really contemplate how the highway system itself affects those other things. So the study reaches its foregone conclusion: the interstate highway system is an end in itself. It must grow, simply because. How else would we avoid congestion? Certainly not by reducing the amount we’re forced to drive for a change. That doesn’t fall within the scope of a report structured only to answer the question, “How much should we spend expanding freeways?”

Ultimately, although the TRB report pays lip service to every cogent critique of 70 years of ruinous transportation policy, it fails to let those critiques become more than side notes. Its underlying premise—that the system needs to grow—can't be allowed to be called into question even by a laundry list of compelling reasons the system shouldn't grow.