Irrigation projects across India are intended to boost agricultural productivity and rural development. In this post, Blakeslee et al. study the long-run effects of access to irrigation on the composition of local economic activity. They find that while access to irrigation has a positive impact on rural villages by increasing agricultural productivity, population density and other indicators of economic development, the reverse is true for irrigated towns, which experience a decline in non-agricultural economic activity.
The diversion of surface water for irrigation has long been one of the most prominent forms through which governments have attempted to exploit natural resources and boost agricultural productivity to foster broad rural development. The impacts of irrigation infrastructure on agricultural production and poverty have been analysed in several papers (starting with the seminal work of Duflo and Pande (2007)). However, it remains unclear whether, in the long run, such a strategy triggers non-agricultural local development, firm creation, and labour reallocation from farm to non-farm activities, or conversely, whether it deepens and perpetuates local dependence on the natural resource and specialisation in agriculture.
We study the long-run effects of improvements in irrigation on the composition of local economic activity in India (Blakeslee et al. 2022). and these have impacted two out of every five villages and towns in the country. We integrate geo-referenced information on the locations served by the irrigation infrastructure with multiple administrative sources to construct a high-resolution dataset at the village- and town-level. This includes data from demographic and economic censuses, satellite observations of dry season cropping from MODIS Enhanced Vegetation Index1, land use and land cover classification indicators from remote sensing (courtesy of the Indian Space Research Organisation), and nighttime lights from the National Oceanic and Atmospheric Administration’s (NOAA) National Geophysical Data Center's Defense Meteorological Program.
Our study
To recover the causal effects of these irrigation projects, we compare settlements that have similar geographic characteristics and are close to, but lie on opposite sides of the projects’ borders. We consider villages and towns that lie within 10 kilometres of the boundary (our results are robust to varying this buffer region between 2 and 30 kilometres). Our approach is similar in spirit to a spatial discontinuity design – the key assumption is that the confounding factors such as geographic characteristics and initial levels of economic development that might impact the economic outcomes we study are continuous at the boundary.
One might be concerned that the locations just inside the project area might have systematically differed from those just outside before the arrival of irrigation. We test this hypothesis using data from the earliest census round (1991) that is publicly available – results from these placebo regressions show that locations on either side of the boundary were similar across a range of economic indicators of prosperity and development prior to canal construction. In other words, only access to irrigation exogenously varies across the command area boundaries in our study sample.
We conduct our analysis separately for three types of units: villages, towns, and regions with multiple locations. This choice is guided by a which indicates that the impact of irrigation depends starkly on the type of region that is hit by the positive, permanent agricultural productivity shock:
(i) In villages (rural locations that are primarily agricultural) population increases relative to the unirrigated villages in the long run. This is because an increase in local wages slows down the outward movement of workers.
(ii) In towns (urban locations specialising in manufacturing), the same shock slows down productivity growth in the manufacturing sector, which in the long run generates a reduction in population and real wages relative to unaffected towns
(iii) In regions with multiple locations (that is, both towns and villages), the model predicts that local aggregate impacts will be negative for large-scale non-agricultural production. The net impact on the population, however, will depend on whether the absorption of workers in villages will dominate the release of workers from the urban location.
In addition to the theoretical motivation for separately examining the impacts of irrigation in rural and urban locations, we empirically verify that towns and villages are remarkably different from each other. Compared to villages, towns are larger in area, more densely populated, and have economies oriented towards non-agricultural production and trade. We also show that there is no differential impact of irrigation on town formation for towns in treatment and control areas of our study sample. Though there is no evidence for endogenous town formation, we conduct additional tests that demonstrate that our main results are robust to accounting for any potential violations of this finding.
Key findings
We first document that irrigation had a positive impact on agricultural productivity in rural villages by allowing them to expand crop production to seasons when it had previously been non-viable. We also show that these areas experienced increases in population density and indicators of economic development (assets and nightlights). The magnitudes of the estimated effects – a 6.1% increase in village population density, a 6.5% increase in light density, and a 3.5% increase in the built-up area – are consistent with those in the existing literature (see Dillon and Fishman (2019) for a review).
However, if we compare irrigated towns with similar neighbouring unirrigated towns, we find the opposite effects to those observed in villages. We observe a 30.8% decline in population density, a 26.1% decline in light density, and a 26.8% decline in the built-up area of irrigated towns. Importantly, towns also experienced a substantial decline in the scale of manufacturing activity and the presence of large firms, as well as a shift in the labour force away from non-agricultural employment.
The heterogeneity in impacts across towns and villages raises the question of what the local aggregate impacts are of the agricultural productivity shocks. Our model suggests that the impacts on the total population will depend on the ratio of towns to villages, and the degree of friction to labour mobility. We estimate this empirically by aggregating outcomes in two ways: first, we define a geographic cell as a town and its surrounding 10 kilometre hinterland; and second, we use 10 square kilometre cells on either side of the command area boundary. We find that, on aggregate, the command area experiences an increase in population density, firm employment, and manufacturing employment; but no change in employment in large firms. However, there are substantial declines for all these outcomes when there is a town present, indicating that losses occurring in towns are not offset by gains to surrounding villages.
It is critical to emphasise that our estimates, whether at the village, town, or cell level, capture the local economic impacts of agricultural productivity gains offered by irrigation. The widespread introduction of irrigation also has general equilibrium country-wide impacts, including the potential acceleration or slowdown in aggregate structural transformation, but these do not lend themselves to causal inference using our approach. The local impacts we estimate occur against that backdrop, and in addition to it.
Conclusion
To sum up, we find that local agricultural productivity gains arising from irrigation expansion can bring substantial benefits to rural farmers, but that they can also potentially hinder local non-agricultural economic activity in relatively more urbanised areas, consistent with findings by Foster and Rosenzweig (2004). We provide evidence that these agricultural productivity shocks have changed the spatial organisation of agriculture, with potentially important implications for aggregate welfare.
Note:
1. MODIS Enhanced Vegetation Index (EVI) shows the density of plant growth, with low values of EVI indicating barren areas.
Further Reading
- Blakeslee, David, Aaditya Dar, Ram Fishman, Samreen Malik, Heitor Pellegrina and Karan Singh (2022), “Irrigation and the Spatial Pattern of Local Economic Development in India”, Journal of Development Economics, 161.
- Dillon, Andrew and Ram Fishman (2019), "Dams: Effects of Hydrological Infrastructure on Development", Annual Review of Resource Economics, 11: 125-148.
- Duflo, Esther and Rohini Pande (2007), “Dams”, The Quarterly Journal of Economics, 122(2): 601-646.
- Foster, Andrew D and Mark R Rosenzweig (2004), “Agricultural Productivity Growth, Rural Economic Diversity, and Economic Reforms: India, 1970-2000”, Economic Development and Cultural Change, 52(3): 509-542.
Comments will be held for moderation. Your contact information will not be made public.