In the first chapter, I estimate an elasticity of irrigation adoption to its gross returns in rural India. Many approaches to estimating this elasticity fail when agents select into adopting irrigation on heterogeneous gross returns and costs. I develop a novel approach to correct for selection using two instrumental variable estimators that can be implemented with aggregate data on gross revenue and adoption of irrigation. I use climate and soil characteristics as an instrument for gross returns to irrigation, and hydrogeology as an instrument for irrigation to correct for selection. I estimate that a 1% increase in the gross returns to irrigation causes a 0.7% increase in adoption of irrigation. I use this elasticity to infer changes in profits from changes in adoption of irrigation caused by shocks to its profitability, and to conduct counterfactuals. First, groundwater depletion from 2000-2010 in northwestern India permanently reduced economic surplus by 1.2% of gross agricultural revenue. Second, I evaluate a policy that optimally reduces relative subsidies for groundwater irrigation in districts with large negative pumping externalities, while holding total subsidies fixed. Under the policy, depletion caused by subsidies decreases by 16%, but farmer surplus increases by only 0.07% of gross agricultural revenue.
In the second chapter, co-authored with Maria Jones, Florence Kondylis, and Jeremy Magruder, we examine the returns to newly-constructed hillside irrigation schemes in Rwanda using a very granular spatial regression discontinuity design. We find that irrigation enables dry season horticultural production which is associated with large increases in labor and input usage and boosts on-farm yields and cash profits by 70%. At the same time, irrigation use remains limited after 4 years. We leverage the spatial discontinuity in access to irrigation to develop a test for separation failures based on farmer behavior on other plots and conclude that separation failures restrain technology adoption. Unlike existing separation tests, our test allows us to distinguish the role of labor constraints from credit and insurance constraints; we find robust evidence that labor constraints limit adoption.
In the third chapter, I develop a new approach to quantify the welfare gains from risky technologies for intertemporal substitution, ranging from agricultural technologies to financial products. Traditionally, these welfare gains are measured either by the technology’s effect on a welfare proxy or by estimating a structural model. Using a welfare proxy may be suboptimal due to noise in measurement and the challenge of converting estimated effects into a money metric, while structural approaches may require strong functional form assumptions and depend on unexpected moments of the data. In contrast, despite some drawbacks, Marshallian consumer surplus is frequently used as a metric for the welfare gains from access to a new product in a static setting, and with sufficient variation in prices may be relatively easy to precisely estimate. I show that under a broad class of models of dynamic optimization which nest Deaton (1991), Marshallian consumer surplus is a reasonable welfare metric for access to an intertemporal substitution technology. I demonstrate how to calculate it, and apply the approach to three experiments which randomly varied either interest rates or prices: I compare the consumer surplus from grants of index insurance in Ghana to their actuarially fair value, I calculate the surplus to households from access to a leading MFI in Mexico, and I bound the foregone consumer surplus due to inattention to the Savers’ Credit among households in the United States. In all cases, the calculation is straightforward, transparent, and can be represented graphically as a “welfare triangle”.
