1. Inter-district price differences are equal to trade costs (in special cases): That is, if a com-
modity can be made in only one district (the ‘origin’) but is consumed in other districts (‘des-
tinations’), then that commodity’s origin-destination price difference is equal to its origin-
destination trade cost. I use this result to infer trade costs (which researchers never fully
observe) by exploiting widely-traded commodities that could only be made in one district.
Using inter-district price differentials, along with a graph theory algorithm embedded in a
non-linear least squares routine, I estimate the trade cost parameters governing traders’ en-
dogenous route decisions on a network of roads, rivers, coasts and railroads. This is a novel
method for inferring trade costs in networked settings. My resulting parameter estimates
reveal that railroads significantly reduced the cost of trading in India.
2. Bilateral trade flows take the ‘gravity equation’ form: That is, holding constant exporter- and
importer-specific effects, bilateral trade costs reduce bilateral trade flows. I find that railroad-
driven reductions in trade costs (estimated in Step 1) increase bilateral trade flows, and show
that the parameters estimated from the gravity equation identify my model.
3. Railroads reduce the responsiveness of prices to local productivity shocks: That is, a district’s
prices are less responsive to its own productivity shocks when it is connected to the railroad
network; however, a district’s prices are more responsive to any other district’s productivity
shocks when these two districts are connected by a railroad line. I find empirical support
for both of these predictions. Specifically, in a novel test for market integration, I find
that railroads caused a dramatic reduction in the responsiveness of prices to local rainfall
shocks, reducing responsiveness to almost zero (even when focusing purely on rainfall vari-
ation across crops, within a district and year). This implies that railroads brought India’s
district economies close to the small open economy limit where local conditions have no effect
on local prices. I also find that a district’s rainfall shocks affect prices in neighboring districts
to which it is connected by the railroad network (to a weak but statistically significant extent).
4. Railroads increase real income levels: That is, when a district is connected to the railroad
network its real income rises; however, improvements in the railroad network that by-pass
a district reduce the district’s real income (a negative spillover effect). Empirically, I find
that own-railroad access raises real income by 18 percent, but a neighbor’s access reduces
real income by 4 percent. However, these are reduced-form estimates that could be due to a
number of mechanisms. A key goal of Step 6 is to assess how much of the reduced-form effect of
railroads can be attributed to gains from trade due to the trade cost reductions found in Step 1.
5. Railroads decrease real income volatility: When a district is connected to the railroad network,
its real income is less responsive to stochastic productivity shocks in the district (which reduces
volatility). Empirically, I find that railroads reduced the responsiveness of real agricultural
income to local rainfall, which suggests a second welfare benefit of transportation infrastruc-
ture (in addition to that found in Step 4) that has not, to my knowledge, been demonstrated
empirically before. However, as with the results in Step 4, a number of mechanisms could
underpin this reduced-form result.
6. There exists a sufficient statistic for the welfare gains from railroads: That is, despite the com-
plexity of the model’s general equilibrium relationships, the impact of the railroad network on
welfare in a district is captured by one variable: the share of that district’s expenditure that
it sources from itself. A prediction similar to this appears in a wide range of trade models
but has not, to my knowledge, been tested before.3 I test this prediction by regressing real
income on this sufficient statistic (as calculated using the model estimated in Steps 1 and 2)
alongside the regressors from Steps 4 and 5 (which capture the reduced-form impact of rail-
roads).4 When I do this, the reduced-form coefficients on railroad access estimated in Steps
4 and 5 fall to a level that is close to zero. This finding provides support for prediction 6 of
the model and suggests that decreased trade costs account for virtually all of the real income
impacts of the Indian railroad network.
Beyond his findings, he uses clever empirical strategy to show that his results are not spurious. In particular, he tests his propositions for areas where railroads were planned put not executed, and finds that they do not get the perceived benefits of the railroads.
While the impact of infrastructure development in the US now would be less spectacular than railroad development in the Indian Raj, Donaldson's paper still has important implications, particularly about the development of freight transport systems and intermodal transportation. The paper is also a terrific read.
No comments:
Post a Comment