Telegraph, Railway Networks, and the Adoption of Agricultural Technology

Governments and international organizations around the world have invested trillions of dollars in transportation and information and communications technology infrastructures. Various prominent models of trade maintain that reducing trade costs and information frictions promotes trade, thereby improving the welfare of the trading partners. Previous studies have found evidence that the expansion of the railroad network in 19th-century India and the connection of the transatlantic telegraph between New York and Liverpool in 1866 increased trade and better integrated the markets (Donaldson and Hornbeck, 2016; Donaldson, 2018; Steinwender, 2018). A broad literature suggests that trade (openness) could further facilitate technology diffusion across regions (Keller, 2004). However, little is known about whether infrastructure can stimulate more profound technology diffusion, and much less is known about whether different types of infrastructure are more effective at facilitating such diffusion.

In this study, I propose to advance the literature on infrastructure, and on technology diffusion by examining the effect of the rapid expansion of railroad and telegraph networks in early 20th-century China (c. 1920s to 1930s) on the spread of high-quality American cotton to 786 rural Chinese counties. I will explore the question in two steps. First, I will estimate the effect of the expansion of railroad and telegraph networks on technology diffusion—specifically the adoption of U.S. cotton—using a differences-in-differences (DID) design to analyze county-level panel data on cotton cultivation. Next, I will examine the causal effect of infrastructure by using the potential railway and telegraph routes as an instrumental variable for the actual railway/telegraph network for a more robust estimation.

Depending on the outcome of the above analysis, I will further examine the influence of two possible channels through which infrastructure could affect technology diffusion. The first channel is that, by reducing trade costs and/or information friction, infrastructure facilitates long-distance trade between remote rural counties and industrialized coastal ports, which gives peasants in these counties a strong incentive to sell lucrative American cotton to coastal textile factories (the "market integration" channel). I will digitalize domestic bilateral trade and price data to empirically test this channel using a county gravity model. The second potential channel is that an integrated market helps transmit the positive productivity shocks from the downstream industry to increasing demands for cotton from the upstream cotton farmers (the "supply chain" channel). I plan to construct a firm-level panel data set of the spinning and weaving industry to empirically test.