Research

"The Impact of Climate Change on Indian Agriculture" (Job Market Paper)
Abstract: This paper estimates the economic impact of climate change on Indian agriculture. I estimate the effect of random year-to-year variation in weather on agricultural output using a 40-year district-level panel data set covering over 200 Indian districts. These panel estimates incorporate farmers' within-year adaptations to annual weather shocks. I argue that these estimates, derived from short-run weather effects, are also relevant for predicting the medium-run economic impact of climate change if farmers are con- strained in their ability to recognize and adapt quickly to changing mean climate. The predicted medium-run impact is negative and statistically significant: I fi…nd that projected climate change over the period 2010-2039 reduces major crop yields by four to eight percent. The long-run (2070-2099) impact is dramatic, reducing yields by roughly 25 percent in the absence of long-run adaptation. These results suggest that climate change is likely to impose significant costs on the Indian economy unless farmers can quickly recognize and adapt to increasing temperatures. Such rapid adaptation may be less plausible in a developing country, where access to information and capital is limited.

“Estimating Quantile Treatment Effects in a Regression Discontinuity Design”
Abstract: Regression discontinuity (RD) is a popular quasi-experimental estimator for mean treatment effects. While the distribution of treatment effects is also of interest, methods for estimating quantile treatment effects have not been developed for RD models. The connection between RD and instrumental variables (IV) noted by Hahn, Todd and van der Klaaw (2001) suggests using quantile IV to estimate quantile treatment effects. This paper develops a simple quantile IV estimator for treatment effects in a RD framework, using the inverse quantile regression method of Chernozukov and Hansen (2006).

“The Short- and Medium-Run Benefits of Sanitation” (joint with Esther Duflo and Michael Greenstone)
Abstract: Poor sanitation and water quality are believed to impose large costs on individuals in developing countries. Due to externalities, sanitation problems cannot be solved at the household level and require group interventions. This paper evaluates the impact of a sanitation program run by an NGO in rural India, which provided toilets, showers and a water tank to over 50 poor villages, on infant and child mortality, adult mortality, school attendance and sanitation-related health conditions such as prevalence of malnutrition, severe diarrhea, typhoid fever, scabies and blindness. Several features of the program make it particularly attractive for evaluation. First, the program required all households in affected villages to participate so the estimated impacts capture all within-village externalities. Second, service was initiated suddenly, unpredictably and at different times for different villages, reducing concerns of confounding with other trends. Finally, data on all outcomes are available beginning several years before the sanitation project came and data collection continues for five years or more after service begins, which makes it possible to assess medium-run impacts.

“Does Public Human Capital Investment Induce Brain Drain? Estimates of the Causal Effect of Education on Migration”
Abstract: A positive association between education and migration is well-documented, but whether this relationship is causal is unknown. I use a plausibly exogenous education intervention from Indonesia’s INPRES school construction program to estimate the causal effect of education on migration. Preliminary estimates indicate that an additional year of primary education increases the propensity to migrate by roughly two to four percentage points. I present a model of public human capital investment and migration choice showing that this causal relationship may reduce the incentive of local governments to invest in human capital if the benefits of such investments accrue to other locales.