Agricultural Externalities in Interconnected Systems: Causal Evidence from Hydrological Networks and Global Commodity Markets
Year: 2026
Type: dissertation
Abstract: Agriculture provides food, fiber, and energy that are fundamental to human well-being. Yet agricultural production and policies can generate externalities, including health risks from chemical use and environmental damages from land-use change. These impacts often extend far beyond the locations where agricultural decisions are made, transmitted through interconnected natural and economic systems. This dissertation examines how hydrological networks and global commodity markets propagate agricultural externalities across space, creating consequences for populations and ecosystems far removed from production and consumption sites. One essay in the dissertation investigates the health impacts of agricultural water pollution by examining glyphosate, the world's most widely used herbicide. Leveraging hydrological flow patterns and variations in upstream glyphosate use across watersheds and time to identify causal effects, I show that glyphosate affects populations far downstream from application sites through waterborne transmission. Using a panel dataset in the US Corn Belt from 2001 to 2014, the results suggest that a 10 kg/km2 increase in upstream glyphosate use increases neonatal mortality by 4.6 percent in lower-income areas, with no observed effects in higher-income regions. The research design also exploits variations in spatial distance, seasonal exposure patterns, and rainfall data to isolate the effects of glyphosate contamination. Additional evidence suggests that water treatment and avoidance behaviors may mitigate exposure risks and contribute to the observed heterogeneity. Another essay examines the global environmental consequences of biomass-based diesel (BBD) policies, highlighting how market-mediated spillovers can undermine sustainability goals. Although biofuels are intended to reduce greenhouse gas emissions from transportation, their climate benefits may be offset by indirect land-use change. Vegetable oils, including palm oil, are the main feedstock of BBD. Combining high-resolution satellite data with a causal econometric framework, this study estimates that global BBD demand from 2002 to 2018 induced the conversion of 1.7 million hectares of forest, primarily natural forests, to oil palm plantations in Indonesia and Malaysia. By modeling the global vegetable oil market, the study further shows that vegetable oils are highly substitutable. Increased demand for non-palm oils in BBD production raises palm oil demand indirectly. This expansion, driven by biomass-based diesel-induced increases in palm oil prices, accounts for approximately one-fifth of regional forest-to-palm conversion during the study period. The resulting land-use change released more than one gigaton of CO2, implying a higher carbon intensity for biomass-based diesel than for fossil diesel. The findings highlight how market linkages across vegetable oils transmit the effects of biofuel policies globally, causing environmental damage far from the regions where biofuels are consumed. Collectively, this dissertation demonstrates that accounting for spatial spillovers transmitted through both natural and economic networks is essential for designing agricultural and environmental policies that fully incorporate their unintended consequences.
Author: Chen, Tzu-Hui J.
Language: English
Subfield: Environmental Engineering
Field: Environmental Science
Domain: Physical Sciences
Keywords: agricultural externalities, glyphosate, agricultural water pollution, neonatal mortality, waterborne transmission, hydrological network, biofuel policy, indirect land use change, palm oil, vegetable oil market, forest conversion, CO2 emissions, international commodity markets, market spillovers, Corn Belt, Indonesia, Malaysia, natural forests, greenhouse gas emissions
Open Access status: green
OpenAlex ID: W7221500795
