Principal Research Question and Key Result Did the railroads built in the Raj reduce trade costs, increase trade flows, and increase welfare in the process? The results of the analysis indicate that…
Theory In a Ricardian trade model with many regions and many commodities trade occurs at a cost (transportation etc.) As different regions have different levels of productivity across products, they have incentives to trade in order to exploit comparative advantage. A new rail link (where the previous transportation options was mainly bullock by road) reduces the bilateral trade cost allowing consumers to buy goods from the cheaper producing district and focus more on producing to their comparative advantage. This model delivers four predictions which are then taken to the data:

  1. If a commodity can be made only in one district but is consumed in other districts then the difference in price for the commodity in those two districts reflects trade costs. This is used to then predict the cost of trade in pre and post railway India.
  2. The gravity equation form of bilateral trade flows indicates that as trading costs decline bilateral trade flows should increase.
  3. Railroads raise real income levels such that when a district is connected to the railroad real incomes should rise. This could occur through a number of channels
  4. It is claimed that the share of autarky (how much a district purchases from itself) is a sufficient statistic for estimating the welfare improvements as being the consequence of lower trading costs.


Motivation Projects aimed at development are often directed at installing infrastructure in order to reduce the costs of trading in order to improve welfare. However, there is a lack of robust empirical analysis that examines the effects of such projects.In terms of the globalization debate, infrastructure can be thought of as one tool for reducing trade barriers. In the case of this paper infrastructure facilitated trade within the country. Often trade debates are about international trade, whereas most trade still occurs within countries and so only to focus on international trade is to miss much of the picture.


Data Data sources are very varied and will be described in the strategy section.
  1. There were several homogenous types of salt produced in Indian districts e.g. Kohat salt which only came from the Kohat mine although it was consumed throughout India. Annual data on 8 types of salt for 124 districts for 70 years is regressed on a variable that lowest cost route effective between the origin of the salt to the destination. This variable is normalized to 1 for railroad, and then values for road, river and coast are estimated using a shortest path algorithm.
  2. Uses the estimates in (1) as for salt, and applies to all commodities.  The regresses the value of exports, on the cost of the lowest cost effective route (the salt estimates). He uses trade flow data with over 1.3 million observations.
  3. Regresses a measure of real agricultural income per acre by district on a rail dummy that equals 1 in all years when some part of the district is connected to the rail network.
  4. A change in autarky in any given district (i.e. how much trade a district does with itself) is driven in this theory by the presence of railroads. If this is the case, then including a measure of autarky in the regression in (3) should induce the coefficient on the rail variable to fall to zero as the effect on agricultural incomes because of rail operates through the openness of a district.


  1. The results indicate that the relative costs of the types of transport being compared to rail are all more expensive per unit distance than rail. In general the elasticity of price to transport distance is 0.135 (1% significance), and the mode specific elasticity is 0.247 (1% significance). Roads appear to be the most expensive presumably as they were the slowest. This is evidence that transportation costs were indeed driving a wedge between the price at origin and the price at destination.
  2. Results indicate that trade flows fall as cost of transport increases. Indeed the coefficient on the log of lowest distance is -1.141 (1% significance).  This is very large. He interacts the distance variable with a variable that measures the value of the cargo and one that measures the weight of the cargo to see if the findings are being driven by some quality of the specific goods being delivered. The coefficient on the distance variable is very similar, and there is no significance on the coefficients on the interaction terms. This is evidence that trade flows increased with decreasing transportation costs.
  3. The log real income increases by 0.164 log points when the dummy equals 1, indicating that agricultural incomes increased when the railroad arrived in the district. This indicates that the impact of trade increasing was welfare increasing. However, this could have occurred through a number of mechanisms, hence why we move on to (4).
  4. When the autarky measure is included the rail coefficient falls to 0.023 and is insignificant, and the autarky coefficient is close to minus 1 and is significant at the 1% level. This is pretty good evidence that the change in agricultural incomes occurred due to changes in openness caused by railroad building.


Robustness In order to get unbiased estimates the railroads need not to have not been built in areas districts that were particularly promising in terms of trade or agricultural development otherwise spurious correlations would be found. In order to test this he uses planned railroad networks proposed by four different organizations/places, and finds no significant relationship between them and agricultural incomes. The Kennedy plan was a particularly interesting test, as it was proposing to build railroads in the most geographically accessible parts which could also have been the parts best suited for trade/agricultural development. 
Problems I’m going to skip this section as the paper is very detailed and I did not really understand all of it. No doubt there are problems. External validity is always an easy one to mention. 
Implications The immediate implications are that connecting areas within a nation can be important for improving welfare by allowing regions to purchase more cheaply what it is costly for them to produce, and to produce more freely what they are better at producing. Under the conditions examined real incomes rose and as such trade could be a good tool in the fight against poverty.Innovation/investment not addressed by this paper.

The context is highly historical and probably not generalizable to the world today. The globalized trading system today could interact with any infrastructure programmes such that we do not see the benefits outlined in this paper. For example, assuming that tariff levels are zero, connecting a port to the interior by rail could reduce the cost of imports such that the interior districts no longer purchase goods from other interior (or similar) districts, thus decreasing incomes in those losing districts (at least in the short term). Perhaps in the long run this exposure to international competitive advantage will increase productivity and efficiency, but it would almost certainly create losses in the short run, and if these occur in an LDC with very low capacity to transition through the production markets, these losses may become structural and permanent.

The paper only looks at agricultural incomes as this was the overridingly dominant form of industry at that time in India. In the modern context it is not clear that we would see similar results in a diverse economy producing a number of different goods.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: