If You Build It, They Will Drive

A reader and transportation advocate in Salem alerted me to a paper that had escaped my attention by Duranton and Turner, two economists from the University of Toronto. It was published in the October 2011 American Economic Review - one of, if not the, pre-eminent journals in economics - and is entitled "The Fundamental Law of Road Congestion: Evidence from US Cities.  In it, the authors try and estimate the capacity elasticity of vehicle kilometers travelled (VKT).  In other words, by how much will vehicle travel increase in response to increases road capacity.  This paper seeks to answer the question: will new road construction ease overall congestion?

[As an aside, how is a paper about driving in the US and published in a US journal allowed to get away with using kilometers instead of miles?? It is an abomination!]

This question is a lot harder to answer than it may appear initially.  You can't just look at road construction and subsequent changes in miles driven (yes, I said miles) because you don't know if the road was built in response to the increase in miles driven, if the increase in miles driven was in response to the new road capacity, or if miles driven would have increased just the same regardless of road construction.  Empirically, this presents a very tricky problem. I won't get into the minutiae of how they addressed this save to say they used historical data on transportation planning to instrument for the road capacity in order to identify the true effect of capacity on VKT.  [Econ students should now be thinking hard about local average treatment effects: the identification comes from things like the original planning document for the interstate highway network in 1947 - how does the variation in planned highways in 1947 affect VKT today?]  The methods are pretty convincing and the results, to my mind, are pretty startling.  Here is the abstract:

We investigate the effect of lane kilometers of roads on vehicle-kilometers traveled (VKT) in US cities. VKT increases proportionately to roadway lane kilometers for interstate highways and probably slightly less rapidly for other types of roads. The sources for this extra VKT are increases in driving by current residents, increases in commercial traffic, and migration. Increasing lane kilometers for one type of road diverts little traffic from other types of road. We find no evidence that the provision of public transportation affects VKT. We conclude that increased provision of roads or public transit is unlikely to relieve congestion.

What this says is that the capacity of elasticity of VKT is one: that for every new lane kilometer of road you build you will get a corresponding increase in traffic - yielding no net benefit to congestion.  Wow.  So opponents of the CRC, for example, might be pleased to note that claims of quicker travel times are probably completely bogus.

But there is more.  They find exactly the same effect for public transportation (and by extrapolation we can pretty safely say the same is true for bike travel) that every car taken off the road by, say, a new MAX route, will be replaced by a new car yielding no net benefit to congestion.

There are a number of pretty serious caveats here though.  They find that was fills in these new roads is a pretty robust increase in commercial truck transport and in the change in individual driving behavior.  This makes sense to this economist - if road congestion is an equilibrium result of the market for driving, if you lower the marginal cost curve, you expect more driving to get back to the point of marginal cost equals marginal benefit (since the benefit hasn't changed).  So if you are a CRC advocate, you can point to the potential increase in commercial truck activity that will result as evidence of increased economic activity in the region in response to the better transportation network.

Interestingly, the increase in highway usage does not appear to come from those switching from city roads to highways - so such new highway construction should not be expected to ease congestion on surface roads.

There is also the fact that this elasticity works both ways.  So it may be that by getting a person to take a bus or ride a bike only convinces someone else to start driving, removing road capacity in order to provide more bike or transportation infrastructure should result in a proportional decrease in overall VKT.

Finally, these results are for highways, but they do have some, arguably less robust results for major roads in urban areas where the elasticity estimates range from 0.67 to 0.89, so high but not 1.  Which means that building out major roads in urban areas should be expected to decrease congestion, but by relatively small amounts.  For example if we use 0.75 as a intermediate number than increasing urban road capacity by 10% should yield only a 7.5% increase in VKT.

So there is lots here for advocates on both sides to wrangle over.  I, being the Scholar, have a duty only to report and interpret.  Go forth and wrangle away...