This is the 17th in our series of posts by job market candidates.
Countries and aid agencies see improving the productivity of agriculture as an important part of development strategy. Farms in the developing world use labor-intensive techniques that rich countries abandoned. For example, farmers in India still use a team of ten women to transplant an acre of rice, a task that farmers in Japan and Italy mechanized over 70 years ago. While the long-run association between mechanization and development is clear, the short-run causal effects of increased mechanization on local economies are poorly understood. Mechanization’s effects depend not only on farmers’ outcomes but on what happens to the displaced workers. I partnered with the government of Telangana, India, to help farmers in 205 randomly selected villages access drum seeders (Figure 1), a $65 device that helps farmers grow rice without manual transplanting. Drum seeders affected not only adopting farmers but also their neighbors through mechanization's effect on labor markets.
Figure 1. Drum seeding technology compared to manual transplanting
Do farmers benefit from drum seeders?
Layering a program where village governments rented drum seeders on top of the government’s existing extension efforts successfully led farmers in treatment villages to adopt drum seeders, especially those with larger land holdings who stood to save the most on labor costs. In addition to renting the devices farmers needed to learn how to adjust their weed and water management. Drum-seeded acreage increased by approximately 40 acres per village (5% of the median village’s total rice area), which corresponds to 400 workdays for women. Farmers in treatment villages hired 7% fewer female workers overall. Despite the reduced labor input, yield and revenues remained unchanged, leading to 6% higher average profits in treatment villages. If the blog ended with these results, this post would mirror the hundreds of agriculture technology success stories discussed on this blog since 2011. However, this study's village-level randomization allows me to unpack the labor market spillovers driving the results.
The labor market effects of mechanization reconcile how modest adoption led to such large changes in average profits for all farmers in villages with drum seeder rentals. Labor costs decreased not only for farmers who use drum seeders but also for those who kept hiring transplanters. Without farming differently, they benefited from their neighbors’ choices. Their neighbors helped solve the problem that inspired the government to promote drum seeders, as transplanting labor markets decongested. Transplanting is a time-dependent task, and when many farmers need to transplant at once, wages can spike. The treatment did not change the median wage transplanters paid, but it reduced the spikes (Figure 2). Overall, farmers who transplanted their rice reported paying workers 6% less per acre of work in villages with drum seeder rentals. These labor cost savings mean that farmers still transplanting in treatment villages now benefit less from adopting drum seeders.
Figure 2. Treatment effect on the distribution of wages farmers paid to transplanters
What happens to workers when farmers start using drum seeders?
The wage effects provide insight into the structure of rural labor markets for rice transplanting, which account for 20% of the crop’s total cultivation cost. Given that labor demand is almost perfectly inelastic in this setting, conditional on the farmer’s choice of planting technology, the wage effects can be interpreted as an estimate of the slope of the labor supply curve. Because rice cultivation is limited by water infrastructure and inflexible land markets, there was no effect on the extent of rice cultivation or output. Land is the limiting factor of production, and transplanting labor is a fixed function of land. The fact there is only a small decrease in the number of women willing to work when wages fall suggests that women have few opportunities beyond transplanting itself. In a phone survey of women who participate in a government workfare program that provides additional days of manual work, I found transplanters in treatment villages wanted more days of workfare. In another project in the same setting exploiting historical variation in land concentration, I find that landed elites suppressed the workfare program during the peak agricultural period (Brownstone and Srivastava 2024 Table 7). Demand for the workfare program, which pays half the median transplanting wage, implies outside options are very limited. In the same phone survey, I found no evidence that the women find jobs outside the village or take up non-agricultural work in the immediate aftermath of mechanization.
In the long run, evidence from Brazil suggests labor substituting agricultural technology is associated with industrial growth (Bustos et al. 2016). In India, evidence from household-level randomization of mechanization access suggests women in mechanizing households substitute work for leisure (Caunedo and Kala 2021). Bridging those papers, I study transplanters whose exposure to the income effects of mechanization varied. The impact of greater farm profits is reflected in women overall in treatment villages reporting their husbands are less likely to let them work outside their village. The impact of lower transplanting wages increasing the price husbands pay for limiting their wives to local work is reflected in the same transplanting subgroup who reported increased demand for government work also reporting their husbands more willing to let them work outside the village. There were no treatment effects on actual work location, and only 2.5% of women reported working outside their village at all. But as mechanization progresses, these shifting attitudes will be important for women’s longer-run response.
How do labor markets and mechanization interact in equilibrium?
In the paper, I also develop a model to capture how labor market structure affects the impact of reducing farmers' barriers to mechanization in equilibrium. The model captures two important dynamics. First, wages affect the extent to which a reduction in the fixed cost of mechanization translates to an increase in the level of mechanization in equilibrium. The first farmers who mechanize, those who have the most land and use the most labor, drive down wages for the remaining farmers, making mechanization less profitable for them. The labor supply elasticity means the same reduction in the fixed cost of mechanization can result in very different equilibrium levels of mechanization. The second dynamic is that the impact of mechanization on aggregate productivity is a function of the jobs displaced workers do. If displaced workers shift to unemployment, as the results of the workfare program suggest, the net impact on aggregate productivity will be zero even if farm profits increase. On the other hand, if women find other opportunities at least as profitable as transplanting, as standard models of structural transformation assume, then the increase in productivity in the farm sector will be the lower bound of the overall productivity increase. The overall interpretation is that mechanization policy is most effective for development when labor for the mechanized task is supplied elastically.
What does this mean for development policy?
Balancing productivity growth with its distributional impacts is a political and ethical challenge. Improving labor market opportunities for women alongside reducing barriers to mechanization for farmers would make the mechanization investments more effective and equitable. For example, schemes like free rural buses for women, which the government of Telangana recently implemented, may have surprising effects on agricultural productivity growth. Alternatively, governments can try to promote mechanization technologies that affect groups, like men, who can adjust more easily to labor market shocks. Finally, firms like Bayer and Shell have large carbon credit projects for direct-seeded rice. Studying how to design credible carbon credits in domains like drum seeding, where farmers' private profits from adoption are meaningful but aggregate adoption is low, is the next step in my research agenda.
Steven Brownstone is a PhD candidate at the University of California, San Diego
Join the Conversation