UC Riverside

This dissertation presents three studies on development economics in India. Each uses quasi-experimental methods to estimate a causal effect that standard measurement misses or that policy design failed to anticipate. The first essay evaluates the Kanyashree Prakalpa (KP), a conditional cash transfer program in West Bengal that pays girls Rs 25,000 to remain unmarried until 18. Using difference-in-differences with NFHS-5 (2019-21) data, I find the program increased marriage likelihood by 4-6 percentage points for girls aged 16-17. Early motherhood rose by 4 percentage points. Education gains were limited to Grades 7-9. KP-exposed brides married better-educated husbands with smaller age gaps, indicating families captured the program's value through improved matches rather than delayed timing. The results show that modest transfers conditional on marital age can backfire when marriage norms are strong and verification is weak. The second essay examines five decades of climate change and technological change in Indian agriculture using district-level panel data from 1966 to 2015 for rice and wheat across 260 districts. Rice loses 9.6 percent and wheat loses 8.9 percent of yields per 100 extreme degree days. Long-difference estimates attenuate toward zero, consistent with adaptation. Green Revolution technologies explain the pattern through crop-specific channels: HYV adoption for rice, irrigation for wheat. A technology that delivered climate resilience for one crop delivered none for the other through the same mechanism. Under a moderate-emissions scenario (SSP2-4.5), projected mid-century rice losses are 2.5 percent, absent further adaptation. The third essay estimates the causal effect of short-run PM2.5 exposure on physical functioning among older Indian adults. I instrument thirty-day PM2.5 with thermal inversion days using WHO SAGE Wave 3 data. A one-standard-deviation increase in PM2.5 reduces grip strength by 1.8 kilograms and walking speed by 0.07 meters per second, each about 8 percent of the sample mean. Effects on self-reports, chronic conditions, and healthcare contact are precisely estimated zeros. Short-run pollution reduces high-effort physical capacity while leaving the rest of the survey unchanged, identifying a health margin between mortality and self-report that prior work does not cover.