Unbundling the process of agricultural technology adoption: Experimental evidence from Kenya
Joint with Shyamal Chowdhury, Jane Mariara, Michael Murigi, Uttam Sharma, Munshi Sulaiman
This paper investigates the process of agricultural technology adoption and diffusion by studying a large-scale agricultural extension program of technology adoption in Central Kenya. We randomize villages to information and goal-setting interventions, and randomly vary the intensity of treatment per village in terms of the share of farmers treated. The training increased technology adoption and farming practices, as well as banana output and profits. The goal setting intervention had no impact beyond that of the training. We observe an inverse-U shaped relationship between the share of households assigned to training, consistent with general equilibrium effects depressing banana prices at high saturation levels. We do not observe statistically significant impacts on higher-level outcomes (profits, income, assets) up to 32 months post-treatment.
Contract structure, time preference, and technology adoption
Joint with Shyamal Chowdhury and Qigang Sun
Do constraints to technology adoption vary by behavioral traits? We randomize 150 villages in Bangladesh into being offered standard microcredit, loans with a grace period, the choice between those two contracts, and control. No discernible average effects are detected on the adoption of mechanized irrigation, hybrid seeds, and chemical fertilizers. However, credit access enhances technology adoption among present-biased farmers, whose output and profits increase. These effects are driven by the standard contract and choice villages, as present-biased farmers select out of the grace period contract. This suggests offering commitment and screening applicants on present bias to enhance agricultural technology adoption.
The encouragement and distortionary effects of conditional cash transfers
Joint with Gharad Bryan, Shyamal Chowdhury, Mushfiq Mobarak, and Melanie Morten
Conditional cash transfer (CCT) programs aim to reduce poverty or advance social goals by encouraging desirable behavior that recipients under-invest in. An unintended consequence of conditionality may be the distortion of recipients' behavior in ways that lower social welfare. We first illustrate instances of distortions arising from CCT programs in the literature. We then concentrate on the case where the CCT causes low return participants to select into a behavior and where social returns and private perceived returns are aligned. In this setting, transfer size plays an important role: the larger the transfer, the stronger the distortion becomes, implying that (i) there is an optimal transfer size for such CCTs, and (ii) unconditional cash transfers (UCTs) may be better than CCTs when the transfer amount is large. We introduce an experimental design that permits a test of this distortionary effect and implement it in a cash transfer program conditional on seasonal labor migration in rural Indonesia. In line with theory, we show that when the transfer size exceeds the amount required for travel expenses, distortionary effects dominate and migration earnings decrease.
Microcredit, financial numeracy, and household financial stress: an empirical investigation
Joint with Isabel Günther
In this paper, we track loan applicants and already-borrowers in urban Uganda to estimate the impact of credit uptake in the first year of borrowing. Our results suggest that there is no average impact on income, assets, food intake, financial distress, or a composite index capturing all four outcomes. However, these average impacts conceal important heterogeneity: borrowers with low financial numeracy levels fare worse as a result of credit uptake, and experience an increase in financial stress. Borrowers with low financial numeracy levels take on loans (with installments) that are larger relative to their income. These findings suggest a role for financial numeracy assessment in the screening of loan applications and extra diligence for loan applicants with low financial numeracy levels.
Nonparametric testing for conditional exogeneity
Joint with Jeffrey S. Racine
This paper reviews testable implications of the unconfoundedness assumption in the treatment effects literature. As there exists no consistent test against arbitrary alternatives of the null of unconfoundedness, it is imperative to test the condition in multiple ways. We propose to assess these testable implications by extending the nonparametric test of equality of conditional densities due to Li, Maasoumi and Racine, [Journal of Econometrics, Vol. 148(2), 2009] to the multivariate case. Importantly, it allows for a mixtures of continuous and categorical variables. Monte Carlo results are presented to illustrate the finite sample properties of this test, and an empirical example on the effects of a job training illustrates its utility.