Network Growth Research Wins Major Award

Sadly, it’s not my research on network growth that won a major award, the penalty of being a civil engineer and thus invisible to economics. But nevertheless congratulations to Stanford economist Dave Donaldson for winning the John Bates Clark medal for work in economic history, and better still, transport history, and even better still for examining network investment and Its consequences.


How large are the benefits of transportation infrastructure projects, and what explains these benefits? To shed new light on these questions, this paper uses archival data from colonial India to investigate the impact of India’s vast railroad network. Guided by four predictions from a general equilibrium trade model, I find that railroads: (1) decreased trade costs and interregional price gaps; (2) increased interregional and international trade; (3) increased real income levels; and (4), that a sufficient statistic for the effect of railroads on welfare in the model (an effect that is purely due to newly exploited gains from trade) accounts for virtually all of the observed reduced-form impact of railroads on real income in the data. I find no spurious effects from over 40,000 km of lines that were approved but – for four different reasons – were never built.

[More discussion of the paper at “A Fine Theorem“]

Hopefully this brings more attention to the subject, which is a vitally important, and bidirectional positive feedback system: the relevant question is not only how does transport affect the economy (Donaldson’s question, as it has been of many before), but also the complementary mutual causality question of how does the economy (including land use) affect the construction of transport. Too bad (according to the NBER version of his paper) he was unaware of the great and long-standing literature in transport geography and regional science, and more recent literature in physics and network science. This starts perhaps with Bill Garrison (as all things do):

  • Garrison, W. L., Berry, B. J. L., Marble, D. F., Nystuen, J. D., & Morrill, R. L. (1969). Studies of highway development and geographic change. Greenwood Press.
  • Garrison, W.L., and Marble, D.F. (1965). “A prolegomenon to the forecasting of transportation development.” Office of Technical Services, United States Department of Commerce, United States Army Aviation Material Labs Technical Report.

See a summary of the literature to 2005 (and it’s probably time for a new synthesis, PhD candidates) in this 2005 paper:

  • Levinson, David (2005) The Evolution of Transport Networks,
    Chapter 11 ( pp 175-188) in Handbook 6: Transport Strategy, Policy and Institutions (David Hensher, ed.) Elsevier, Oxford

Obviously it’s an area I have been researching (with both empirical and simulation methods) for a number of years, hopefully making what I think are useful scientific contributions. For some more recent examples, my 2008 paper Density and Dispersion and the papers collected in 2011’s Evolving Transportation Networks by Feng Xie and myself from his MS Thesis and Dissertation, among others. Other of my publications, including the work of Feng Xie,  on the subject are listed at the bottom of this post. I will also note a couple of Special Issues that I edited on the subject:

Network Growth, My Contributions

The development of transportation networks is a function of policy, planning, and engineering decisions, the inherent geographical and topological structure of networks, and traveler preferences dictating how the network is used. Research into Network Growth aims to disentangle these phenomena, and by doing so, understand the implications of present decisions on future options.

My first major contribution was the discovery of the self-organizing nature of the hierarchy of roads. Some roads are more important (carry more traffic at higher speeds) than others (e.g. major highways vs. local streets). Current design guidelines suggest how a road hierarchy should be laid out by highway planners. However, even in the absence of planners, using a new agent-based simulation model describing the actions of travelers and investment in the network, the interaction of travel behavior (travelers seeking the shortest time path) and simple feedback rules (resources being spent to improve links in proportion to traffic) will produce a hierarchy of roads very similar to what is observed in practice, even if roads start from an undifferentiated state aside from their spatial position (Yerra and Levinson 2005, Levinson and Yerra 2006). We showed that network structure (and the network growth rules) can affect network reliability (Zhang and Levinson 2008), an overly hierarchical structure had serious reliability problems, while the grid network had better efficiency performance, as well as error and attack tolerance.

The second major contribution was the establishment from empirical findings that transportation and land use co-evolve: certain transportation investments are positively and significantly related to future land development, and that land development is positively and significantly related to subsequent transportation investment. Using data compiled from the UK census and the historical development of the surface and underground railway systems in London, a 17 decade time series for 33 boroughs of London was used to rigorously test the hypotheses (Levinson 2008a), which corroborates earlier findings based on five decades of highway and land use data in the Twin Cities (Levinson and Chen 2006). This positive feedback system is convergent, and as systems mature (e.g. the transportation network is built out) its subsequent effects are steadily weaker.

The third major contribution from this research is demonstration using analytical models that governance is endogenous to the transportation technology investment process. As transportation networks get faster (either through investment using existing technologies, or through development of faster new technologies), the amount of inter-jurisdictional traffic increases, and these positive spillovers create demand for higher levels of government (state v. county, e.g.) to manage transportation funding. (Xie and Levinson 2009 “Governance Choice on a Serial Network”). This comports with the historical evidence that transportation has become increasingly dependent on higher levels of government for funding as longer distance (e.g. interstate) travel takes a large share of the total. These three contributions are detailed in the book Evolving Transportation Networks (Xie and Levinson 2011)


References for my work on network growth and evolution, and its causes and consequences.

I know, these are specialist field journals, not Nature and Science, and so get little visibility. But they should all surface in a search of Google Scholar with little difficulty. We of course were unaware of his more recent work, which was in NBER, and will be buried in the obscure economics literature (Supposedly it’s forthcoming in AER, some podunk field journal in some social science sub-discipline). There is a vast chasm between fields.