Evidence from canals in India shows that labour mobility is a key adjustment channel to agricultural growth
The structural transformation of an economy out of agriculture is perhaps the most universal fact of development. Every country in the world that has experienced substantial economic growth has also seen a drastic fall in the share of its workers in agriculture.
The mechanism for this pattern has been of interest to development economists for generations (Schulz 1953, Lewis 1954, Johnston and Mellor 1961). Structural transformation and development surely move together, but the direction of causality and geographic scale at which it operates are still deep questions in the field. An improved understanding of this relationship is highly relevant for policy as well — do poorer countries need to focus on improving productivity in agriculture, trusting that advanced industries will follow? Or conversely, does agricultural productivity change follow development in the other sectors, motivating emphasis on the manufacturing and services sectors? Will investments in agriculture cause villages to become centers for microenterprise and larger-scale industry?
Studying India’s canal network
In a new working paper (Asher, Campion, Gollin and Novosad 2022), we study one of the largest agricultural investments in human history — India’s vast canal network. Built through the 19th and 20th centuries, India’s canals stretch some 300,000 km and deliver water to over 100,000 villages. Access to canal irrigation gives farmers a reliable water source, which is particularly valuable in the winter cropping season when rains are scarce.
A challenge in studying the effects of agricultural productivity change in the long run is that agricultural technology tends to diffuse across space. As high-yield seeds or chemical fertilisers prove their worth, more and more farmers adopt them, making it harder to find a ‘control’ group, especially in the long run. Canals, on the other hand, are fixed in space — they only ever serve a particular area, making it possible to clearly identify ‘treated’ and ‘untreated’ places decades and even centuries after their construction.
Additionally, because canals in India were mostly built long ago (most canals in our sample were completed before 1980), the effects we observe today are long-term: they reflect the equilibrium outcome of a permanent shock to agricultural productivity. They thus provide us a unique opportunity to study how agricultural productivity change affects the spatial distribution of both people and firms in the long run, after both have had plenty of time to relocate.
Indeed, the movement of people into canal regions was a central issue faced by the British administrators who oversaw the construction of canals across the subcontinent in the 19th and first half of the 20th century. Maclean (1885) attributes a doubling of the population and the “most prosperous condition” of agriculture and commerce in the Godavari valley to the “extension of irrigation by canals.” The Punjab Canal Colonies attracted over 1 million migrants to newly arable lands, leading colonial administrator James Douie to state that “[settlement] is the handmaiden of irrigation” (Douie 1914).
Identifying the effects of irrigation from canals
To study the long-run effects of canals, we matched geospatial data on every medium and large canal in India to a wide range of administrative and satellite data at high geographic resolution. We use the gravity-driven nature of canals to identify the direct effect of irrigation: since water only flows downhill, farms below the canal are much more likely to benefit from the irrigation network. This motivates a regression discontinuity analysis — comparing places with elevations just above the canal to places with elevations just below. We restrict the sample to villages with otherwise highly similar geographic characteristics, to ensure we are not comparing rugged mountainsides to fertile valleys.
Sure enough, below-canal villages today report substantially higher canal irrigation and higher agricultural yields (proxied by a satellite vegetation index); the natural experiment holds up and canal-irrigated villages have sharply higher agricultural productivity, even many decades after canal construction.
Figure 1 Agricultural outcomes from different types of irrigation
Note: The figure shows regression discontinuity results for main agricultural outcomes. Blue points indicate normalised treatment effects that are statistically different from 0, and grey dots indicate insignificant results.
Interestingly, these canal-irrigated villages show no sign of structural change; we estimate a precise zero effect of canals on the non-farm share of the labour force. However, they do have substantially higher population, and specifically, a higher landless population. Only landowners have higher living standards in irrigated areas. The results are similar in an analysis of spillovers that compares the above-canal regions to geographically similar places further away: above-canal villages do not seem to be experiencing substantial spillovers in either population, the farm sector, or the non-farm sector.
Figure 2 Non-farm outcomes
Note: The figure shows regression discontinuity results for the main non-farm outcomes. Blue points indicate normalised treatment effects that are statistically distinguishable from 0, and grey dots indicate insignificant results.
Labour mobility in the long run may explain productivity gains
Long-run labour mobility is the best explanation for our findings. Over the long run, people are mobile, especially landless labourers. Even a small rate of internal migration can result in substantial population flows over the decades-long period that we study. Canal irrigation raises agricultural productivity, initially putting upward pressure on agricultural wages. Over the long run, new workers move into irrigated regions until wages are again equalised with the broader economy. Land, in contrast to labour, is a fixed factor — you can extend the agricultural area only marginally — and so landowners get persistent benefits.
However, this is not the end of the story. In a difference-in-difference analysis, we study whether towns grew more in canal regions in the decade after canals were built. To do this, we take advantage of a long-run time series of town populations going back to 1911. Unfortunately there is no long-run time series data on any of our other outcomes. We find that towns in proximity to canals grew substantially faster following canal completion.
The magnitudes of the population movements implied by our estimates are striking. India’s canals have drawn almost 45 million additional people to canal regions — about 5 million people to regional towns and 40 million people to villages directly irrigated by canals. In the end, this can account for a substantial degree of structural change: a back-of-the-envelope calculation suggests that irrigation canals could account for 3-5% of India’s change in urbanisation since 1950.
Population flows are key to structural transformation
Did canal irrigation cause structural transformation in India? Our research suggests that it did, in a big way. But it did so not by bringing firms and factories to the rural places experiencing agricultural productivity gains, but instead by drawing more workers into the urban spaces nearby.
Our findings highlight that population flows are central to long-run development. Everywhere in the world, people move to opportunity. Long-run migration substantially shapes and mediates the effects of place-based investments. Structural transformation is not just a story of occupational change within locations across sectors, but also one of mobility across space.
Editors' note: This column is published in collaboration with Ideas for India. This column draws on research funded by STEG.
References
Asher, S, A Campion, D Gollin and P Novosad (2022), "The Long-run Development Impacts of Agricultural Productivity Gains: Evidence from Irrigation Canals in India", Working paper.
Douie, J M (1914), "The Punjab Canal Colonies", Journal of the Royal Society of Arts 62(3210): 611-623.
Maclean, C D (ed.) (1885), Manual Of The Administration Of The Madras Presidency Vol.1.