UC Berkeley

This dissertation examines how government policies, international trade shocks, and climate variability affect environmental and agricultural outcomes. It draws on large-scale datasets constructed from satellite imagery, administrative records, and market data, and combines causal inference methods with applied machine learning to quantify these effects. It evaluates the effectiveness and welfare implications of conservation subsidies in reducing water pollution, quantifies the impact of international trade shocks and government compensation programs on agricultural commodity prices, and analyzes how climate variability affects carbon use efficiency and crop yields across the United States.The first chapter of the dissertation estimates the effectiveness of the U.S. Department of Agriculture's Environmental Quality Incentives Program (EQIP) cover crop subsidies to mitigate water pollution from agricultural runoff and leaching. Using a novel field-level dataset on cover crop adoption from 2007 to 2023 constructed from satellite imagery and machine learning, combined with water quality monitoring and policy data, the analysis exploits the staggered rollout of the USDA's Mississippi River Basin Healthy Watersheds Initiative (MRBI) for causal identification. A difference-in-differences event-study design shows that increased subsidies raise cover crop adoption by 1.75 percentage points, corresponding to a 34 percent increase relative to baseline, with effects that persist for over a decade but exhibit substantial inframarginal participation. Linking upstream adoption to downstream water quality outcomes, the results indicate that a one-percentage-point increase in adoption reduces total nitrogen concentrations by 0.83 percent. A cost-benefit analysis implies a benefit-cost ratio of approximately 2.52 for the median polluted sub watershed, indicating positive net social returns. These findings provide large-scale causal evidence that voluntary conservation programs can generate meaningful environmental benefits, while highlighting the importance of targeting to improve policy efficiency.The second chapter studies the short-term impact of the U.S.-China trade war on U.S. agricultural commodity futures prices. The trade conflict imposed substantial retaliatory tariffs on key U.S. agricultural exports and was accompanied by large-scale government compensation through the Market Facilitation Programs (MFP). This chapter examines how tariff increases and relief payments affected futures prices, which play a central role in farmers' income expectations and production decisions. The analysis uses daily futures price data for major grains from 2004 to 2020, combined with information on tariffs, government payments, agricultural fundamentals, and weather conditions. To address non-stationarity, all variables are specified in first differences. The results show that tariffs had large negative effects on soybean prices, with a 25% tariff increase leading to a 25.5% decline upon announcement and a 17% decline upon implementation. Wheat prices also declined, while other commodities exhibited limited responses. MFP payments had offsetting effects over time, initially supporting prices in 2018 but contributing to price declines in 2019, suggesting distortions in expectations and production decisions.The third chapter studies the trends and climate responses of carbon use efficiency (CUE) and corn yield across the contiguous United States. This chapter integrates satellite-derived measures of net primary productivity and gross primary productivity from the Moderate Resolution Imaging Spectroradiometer with climate variables from the European Centre for Medium-Range Weather Forecasts. Using these data, the study constructs CUE and standard climate metrics and examine their spatial heterogeneity. The analysis employs a random forest algorithm to identify key climate drivers of CUE and crop yield, and estimate their responses to climate variability using a spatial moving window regression approach. The results show that growing degree days (GDD) have the highest predictive power for both CUE and yield, while extreme degree days (EDD) are the least important. Yield generally increases or remains stable with higher GDD and precipitation, whereas CUE declines with higher GDD in northern regions and exhibits more heterogeneous patterns in the south. Notably, there are some exceptions where yield is negatively correlated with precipitation in the Missouri and Mississippi River Valleys. These findings suggest that, under continued warming, CUE is likely to decline across much of the northern United States, even as yields remain stable or increase. The chapter also introduces an open-source package that facilitates automated retrieval of remote sensing data through Google Earth Engine and county-level agricultural production data from the USDA via an application programming interface.Together, the chapters provide new evidence on the impacts of agricultural policy and climate change, highlighting the role of policy design and behavioral responses in shaping both economic and environmental outcomes. In doing so, this work contributes to the design of agricultural and environmental policies toward economic efficiency and sustainability.