2026 Theses Doctoral

Essays in Climate and Environmental Economics

Zuniga Valladares, Roberto Alfredo

This dissertation comprises three essays in climate and environmental economics, each examining how a changing climate interacts with economic decisions and institutions in the economy. The essays study, respectively, whether subsidized rural credit builds or undermines farmers' climate resilience in Brazil; how anticipated climate change shapes global commodity price volatility and the role of storage as an adaptation mechanism; and how the spatial composition of global warming shapes estimates of its macroeconomic damages.

Chapter 1 asks whether subsidized agricultural credit—widely justified by the goal of building resilience to drought, heat, and other weather shocks—actually does so, using Brazil's main family-farming credit program (PRONAF) and farm-level soybean yields constructed from satellite imagery for the 2013–2024 crop years. Exploiting the program's four-fiscal-module eligibility rule, I bring three designs to bear: a fuzzy regression discontinuity in farm size, a spatial discontinuity across municipal borders, and within-property panel regressions that interact credit with insurance type. Marginal access to credit does not raise soybean yields—both discontinuity designs place the effect at or below zero—and eligible farms are more sensitive to extreme heat, not less. This added heat sensitivity is concentrated among borrowers whose loans are bundled with Proagro, a credit-linked weather guarantee. The pattern reflects not credit steering soybean onto climatically marginal land, but a shift toward a heat-sensitive crop combined with insurance that weakens the incentive to protect it. I conclude that, delivered through this institutional design, subsidized credit can finance maladaptation, understood as potentially socially-inefficient risk exposure, rather than adaptation.

Chapter 2 (joint with Nam Tran, Wolfram Schlenker, and Michael Roberts) studies how anticipated climate change affects global food price stability. Climate change is projected to make crop yields more volatile, yet higher storage levels have the potential to buffer against this volatility. We extend the standard competitive storage model in two ways: we relax the canonical assumption of i.i.d. weather-induced yield shocks, allowing rational market participants to anticipate that the yield distribution itself is evolving over time; and we allow for a supply response to long-term price movements. We project that average global yields fall 10% by 2100 while their coefficient of variation nearly doubles. Market-driven storage adaptation absorbs much of the resulting risk: storage levels increase by 150%, dampening price fluctuations through intertemporal arbitrage. Even so, price volatility rises by 75% and large price spikes become roughly 50% more frequent. Absent storage adaptation, price volatility would double and large spikes would be three times more frequent. The results underscore the importance of storage infrastructure and global market integration in buffering food systems against climate-induced production risk.

Chapter 3 asks whether standard estimates of climate damages, which ignore the spatial composition of warming, understate the costs of sustained temperature rise. Existing approaches rely either on local weather variation, which misses global spillovers, or on shocks to global mean temperature, which lump together distinct climate phenomena—such as El Niño and greenhouse-gas-driven warming—into a single measure. I decompose the global temperature field into distinct spatial patterns using principal component analysis and use the resulting indices as separate shocks in a structural vector autoregression. The leading pattern isolates a warming signal consistent with the geographic fingerprint of human-caused climate change, while subsequent patterns capture recognized climate oscillations. Shocks to this forced-warming pattern produce persistent declines in GDP per capita of 20 to 25 percent per degree of global warming over a ten-year horizon—larger and more persistent than shocks to global mean temperature, whose effects fade at longer horizons. Projecting damages under the SSP2-4.5 scenario, this approach better reproduces the spatial distribution of projected warming and yields end-of-century losses of 50 to 60 percent of GDP per capita, with tighter confidence intervals than estimates based on global mean temperature alone. These results suggest that the composition of global temperature variation is quantitatively important for climate damage estimation.