The effects of transportation infrastructure on cities: A review of the evidence

We would like to understand the effect that transportation investments have on the way an economy operates, on the welfare of its residents, and on the environment. Ideally, our understanding is sufficient to be of use to the World Bank, where they must confront questions like "How many roads should we build in Kuala Lampur, and how should the network be arranged?" We can imagine two classes of answers to this question. The first is positive. We can try to make statements like "Adding 1km of roads will cause 10 jobs to be created and 100 more km of driving per year". This class of answers allows us to assess whether infrastructure will achieve particular policy goals, and also to compare the cost of achieving policy goals across different policies. If we know that adding a road creates one in a city we can compare the cost of this road with an increment to educational attainment of tax policy which also creates one job. The second class of answers is explicitly normative. We can try to make statements like "Adding 1km of roads will be welfare increasing". The current state of knowledge allows substantive positive statements about the effects of transportation investment and also an assessment of the cost effectiveness of these policies. We have a very limited ability to make absolute welfare statements about the effects of transportation infrastructure, but do somewhat better at ranking projects according to their effects on welfare. Our understanding of the effects of public transit is dramatically inferior to our understanding of the effects of roads. In the abstract, the problem of understanding the effects of transportation investments can be resolved easily. We would consider a set of identical cities, treat half of them with roads systems, designing the systems carefully so that the sample variation allowed us to think about the implications of different extents and network designs, and watch what happened over the next two or three generations. Given such an experiment, it would fairly straightforward to resolve the questions I pose above. While one can just barely imagine the World Bank ‘treating’ a random sample of cities with subways or similar investments, we probably should not wait until the results of this experiment are available until we make out infrastructure spending decisions. Thus, we are left to make what inference we can by observing the world around us, generated as it is by an equilibrium rather than an experimental process. Such data poses at least three important problems. First, transportation investments, roads and public transit, are often very long lasting: Rue St. Jaques in Paris was built by the Romans, the US interstate System has so far lasted several decades. This means that an evaluation of these investments needs to be done over generational time scales in which land use, trade patterns, the locations of populations, and production patterns can adjust to the new infrastructure.