Abstract: Cooling demand is rising rapidly in the Global South. This paper examines the role of electricity prices in driving this demand by studying the consumer response to a large electricity subsidy in an Indian state. Under this program, residential consumers with utilization below 10 kWh per day are exempt from paying their bill, whereas consumers that exceed this threshold pay for all units consumed. Using billing data from the state utility, I estimate changes in the temperature-response function following the implementation of the subsidy. I find that nonlinear pricing distorts cooling demand: consumers initially below the threshold increase cooling and consumers initially above the threshold suppress cooling to retain the subsidy benefit. Cooling demand suppression disappears when consumers are exposed to high heat, suggesting temperature mediates the price elasticity for cooling. Lastly, I use merged household survey data to capture responses in the extensive margin, revealing that air conditioner adoption increases following the subsidy implementation, driven disproportionately by lower-income households. Together, these results illustrate that electricity prices shape the demand for cooling and can serve as an important tool for climate adaptation policy.

2026 Scheduled Presentations: LSE Environment Week (London), USAEE North American Conference (Pittsburgh), NEUDC (Cornell), SEA Conference (Houston)

Previous Presentations: SEA Conference 2025, APPAM Conference 2024, MSU Energy and Environment Day 2024

Abstract: Variation in government performance is well documented across countries but much less is known about variation at finer geographies or its proximate causes. Using both observational and experimental data from a single Indian state, we document substantial variation in the delivery of key citizen services and decompose it into contributions from managers and locations. First, bias-corrected AKM (Abowd et al., 1999) variance decompositions show that location effects are substantially larger than manager effects: location variance is nearly five times manager variance and accounts for over half of variation in service delivery time.  Second, a movers design reveals that destination office productivity remains unchanged when managers transfer in. Consistent with this, support-staff productivity is invariant to the identity of the supervising manager, suggesting that relatively immobile support staff are a key location-specific factor. Finally, a state-wide RCT that permanently tightened delivery deadlines shows that the productivity consequences of these location-based constraints can be mitigated by binding regulatory mandates: low-performing offices--where pre-existing deadlines were binding--substantially improved productivity, regardless of manager quality. Taken together, our results suggest an important role for location-based factors in service delivery.

Previous Presentations: Advances with Field Experiments 2026 (University of Chicago), NBER Summer Institute - Development Economics 2026*, NEUDC 2025 (Tufts), State Building in Developing Countries at UC Berkeley 2025

* presented by co-author

Abstract: In 2009 the Philippines introduced a mother tongue-based multilingual education language policy requiring the “mother tongue” as the language of instruction (LOI) in kindergarten through grade 3. Using teacher classroom language data collected from four LOI groups in 2019, we compared the frequency of teachers’ use of the target LOI in different contexts, including urban versus rural classrooms, classrooms with relatively homogeneous student language backgrounds versus more heterogeneous classrooms, and classrooms with materials in the target language versus classrooms without. We also examined language usage against characteristics of the teacher populations, including language background, years of experience, training, and beliefs about the best language for initial literacy. The results strongly suggest that the most influential levers for increasing teacher usage of a designated LOI in these contexts are ensuring that teachers are assigned to schools where the LOI matches their own first language and providing teaching and learning materials in the target LOI, especially teacher’s guides. These two factors were more strongly and more consistently correlated with teacher use of the LOI than all other variables examined. The linguistic homogeneity of the student population also showed a statistically significant though lower impact on teacher language usage.

Abstract: School-related gender-based violence (SRGBV) remains difficult to measure because of high sensitivity and response bias. However, most SRGBV measurement relies on face-to-face (FTF) survey administration, which is susceptible to increased social desirability bias. Widely used in research on sensitive topics, Audio Computer-Assisted Self-Interview (ACASI) allows subjects to respond to pre-recorded questions on a computerized device, providing respondents with privacy and confidentiality. This brief contains the findings from a large-scale study conducted in Uganda in 2019 where primary grade 3 students were randomly selected to complete surveys using either ACASI or FTF administration. The surveys covered school climate, gender attitudes, social-emotional learning, and experiences of SRGBV. Through this study, we find that although most survey responses were comparable between ACASI and FTF groups, the reporting of experiences of sexual violence differed drastically: 43% of students in the FTF group versus 77% of students in the ACASI group reported experiencing sexual violence in the past school term. We also find that factor structures are similar for data collected with ACASI compared with data collected FTF, though there is weaker evidence for construct validity for both administration modes. We conclude that ACASI is a valuable tool in measuring sensitive sub-topics of SRGBV and should be utilized over FTF administration, although further psychometric testing of these surveys is recommended.