Levinson and Xie on First Mover Advantages in Networks
It took me longer than expected to get to this manuscript, but this paper appears in the same volume as the David King piece I wrote about here.
First off, Levinson is one of the creators of the Journal of Transportation and Land Use, and when I squinted at the length of the manuscript, I was thinking rather that, given an editors’ discretion, the piece had just gotten too long–it’s over 20 single-spaced pages, which is a lot for a single journal article. However, I am often put off by how little appears in a given paper, and this manuscript just has a ton of interesting material in it, and it’s nice to see all the ideas laid out, comprehensively, in one place. I wish other journals would calm down about their submission length enough to let authors really cover complex and innovative ideas like this. In general, with shorter papers, the literature review is what suffers as that’s where authors cut down. Here, you get to see the rich body of theory that the authors are drawing from.
With that out of the way, here is the full citation, and the paper is freely available to read on the web
Levinson, D. & Xie, F., 2011, Does First Last? The existence and extent of first mover advantages on spatial networks [pdf download], Journal of Transportation and Land Use. Vol 4 Issue 2 [Summer 2011] pp. 47–69 doi: 10.5198/jtlu.v4i2.197
From their abstract:
This paper examines the nature of first-mover advantages (FMA) in the deployment of spatially differentiated surface transport networks. A number of factors explaining the existence of first-mover advantages have been identified in previous research on market advantages to first-mover businesses. However, the questions of whether these factors exist in spatial networks, and of how they play out with true capital immobility have remained unanswered. By examining empirical examples of commuter rail and the Underground in London, frst-mover advantage is observed and its sources explored. A model of network diffusion is then constructed to replicate the growth of surface transport networks, making it possible to analyze first-mover advantage in a controlled environment. Simulation experiments are conducted, and Spearman rank correlation tests reveal that first-mover advantages can exist in a surface transport network and can become increasingly prominent as the network expands. In addition, the analysis discloses that the extent of first-mover advantages may relate to the initial land use distribution and network redundancy. The sensitivity of simulation results to model parameters are also examined.
When they trace the development of the London underground over time, their point becomes clear: the first stations always serve more passengers than later ones, but the relationship is not so mechanistic as to predict ridership completely (and you wouldn’t assume that anyway). They dig down to find that the source of the first mover advantage isn’t really first being first: it’s related to what they call spatio-temporal location:i.e., the number of connections (positive correlation) and the travel time (negative correlations). So there is probably some mutually reinforcing relationships between station placement (we place stations where they will serve the most demand at the beginning) and then those stations alter demand in ways that reinforce the first stations’ advantages.
With airlines, the case is a lot messier and weaker, since many places have only one airport. They look at six cases involving persistence at airlines’ first market areas, and show that with the exception of one case, airlines continue to dominate in their original market areas. I had trouble following the logic in this section because, as the authors note, airlines can move their equipment in and out of airports, while airports are where they are. I’m troubled somewhat by the same concerns on the container port examples: what emerges from these two sections appears to be FMA for market penetration rather than a FMA that is inherently tied to the physical geography and subsequent development of the rest of the network–though port geographies are more inherently crucial to subsequent market stories than anything else. The goal of these historical explorations is to try to get us to a generalizable set of principles for first mover advantages for spatial networks–an ambitious goal for development cases like airports and ports where the actual facilities are relatively sparse and given the idiosyncratic way in which airport and port development was governed in the US. Perhaps there is an international example that could enhance the story with fewer institutional and regulatory interventions muddying the waters.
The road links exemplar is much more analogous to the rail stations exemplar and makes the authors’ case much more readily. In any case, it is really nice to learn more about the timeline of these facilities.
That part of the manuscript is just a warm-up. In the rest of the manuscript, they develop a place/link formulation model that functions via two-loop iteration. The place formulation model predicts a point in space-time where new places emerge (places being a set of land uses). In the model, they have housing and job locations, and those emerge at the center of the cell clusters representing land development in the model. Those centers or centroids emerge in space as places where the accessibility is better in relation to others–local peaks are those that have greater accessibility than all its neighbors. The local peaks are not necessarily places yet in the model: these are locations where the potential exists for place development because of the accessibility enjoyed from that location.
Whether a place formulates from those local peaks depends, (remember this is iterative) again on the local peaks’ accessibility relative to others’ potential. I assume that a hierarchy emerges among this set based on the first selected. From there, a link formulation and simple travel demand models develop the network between places.
The authors then create two experiments: in A, the numbers of jobs and workers are randomly placed throughout the cells contained in the geography. In B, there is a decay factor (beta1) that describes the decrease in jobs with distance from the center place and a growth factor (beta2) that describes the increase in workers with distance from the center (Introducing an a priori need to travel). They then test
1) whether FMA manifests and then becomes reinforced over subsequent iterations
2) whether FMA is more important in Experiment B; and
3) whether FMA declines when duplications appears in the network (more than one route becomes available).
In the simulations, they find evidence of FMA advantage, particularly in Experiment B, which more closely reflects how urban lands really develop. They also find evidence that the FMA becomes reinforced over time, and that it becomes more reinforced in Experiment B.