Research
Job Market Paper
When Pollution Rises, Reporting Falls: Coal Mining and Self-Monitoring under the Safe Drinking Water Act
Abstract
One regulatory model requires firms to self-report their own compliance, placing the cost of monitoring on the firm and creating an incentive to shirk on reporting. That incentive grows as monitoring costs rise, so if costs rise precisely in states of regulatory concern, firms under-report exactly when reporting is most valuable to the regulator. I document this perverse incentive among drinking water utilities in the United States. Coal mining is a source of water contamination, and sulfur emission regulation under the Acid Rain Program (ARP) drove exogenous shifts in the geography of coal extraction, altering contamination from upstream mining and raising the costs of both self-reporting and treating. To measure utility exposure to mining, I adapt an instrument for coal production based on the reallocation of coal supply across high- and low-sulfur mines following the ARP. Utilities with more upstream mining show higher arsenic, selenium, and barium concentrations, yet no increase in recorded contaminant limit violations. Instead, self-reporting of water quality falls by seven to ten percentage points, even as reporting requirements increase with contamination. Regulators conduct sanitary visits to substitute for utilities’ under-reporting: an additional upstream mine raises the likelihood of a regulator visit by 14 percentage points, yet those visits do not uncover contaminant limit violations. Sanctions do not rise to deter under-reporting; an additional upstream mine reduces the likelihood of formal enforcement by 5.7 percentage points. Regulators could therefore increase self-reporting by increasing penalties for under-reporting.Work in Progress
A More Complete Valuation of US Lake Water Quality (with Rouzbeh Rahai)
Abstract
Standard hedonic models of water quality capture only the value of lakes near a home, understating willingness to pay for water bodies that residents visit but do not live beside. This paper estimates the marginal willingness to pay for lake water quality across more than 300 Core Based Statistical Areas in the contiguous United States using an embedded recreation demand model within a hedonic house price framework. In a first stage, I estimate a discrete-choice recreation demand model at the census block group level, in which residents choose among lake recreation sites within their local market or an outside option, as a function of travel cost and site cyanobacterial harmful algal bloom (cyanoHAB) levels measured from daily satellite imagery. This yields a block-group-level expected compensating surplus from lake recreation. In a second stage, this recreation surplus is embedded as a regressor in a hedonic house price model, alongside local dissolved oxygen readings, to recover the total (direct amenity plus recreational) value of lake water quality. The analysis draws on three novel data sources at national scale: cell phone mobility data (Advan) tracking approximately 46 million devices to measure recreational visits to over 2,000 lakes, satellite-derived cyanobacterial cell counts (NOAA’s Cyanobacterial Assessment Network) as a high-frequency, highly visible measure of water quality, and ATTOM property transaction records for single-family homes sold between 2019 and 2022. This is the first application of national cell phone mobility data within an embedded recreation hedonic model, and the largest-scale hedonic estimate of lake water quality to combine recreation demand and property values. By capturing the value that non-shoreline residents place on lakes they travel to visit, the approach offers a more complete accounting of the benefits of pollution abatement policy than hedonic models restricted to properties near the water’s edge.Are Municipal Heat Ordinances Effective Climate Adaptation Tools? (with Sheila Olmstead, Mary Evans, and Arnab Ghosh)
Abstract
Rising extreme heat from climate change poses growing risks to worker health, particularly in outdoor sectors like construction, yet the United States has no federal workplace heat standard and protections vary widely by state and locality. Texas offers a rare natural experiment for evaluating the causal effects of municipal heat protection ordinances: Austin (2010) and Dallas (2015) enacted local worker heat protections, and in September 2023 the Texas Legislature’s HB 2127 abruptly nullified these ordinances statewide and preempted future municipal action. We combine 2010–2026 Texas Workers’ Compensation claims data (38 million individual claims, including heat-related illnesses and workplace injuries potentially linked to heat via cognitive and physical effects) with 4km gridded PRISM daily temperature, precipitation, and humidity data aggregated to commuting zones, and plan to incorporate construction-project-level daily productivity records from TXDOT Daily Work Reports. Using a binned-temperature difference-in-differences and event-study design that exploits both the staggered adoption of the two city ordinances and their simultaneous statewide repeal, we estimate (1) the impact of municipal heat ordinances on medical utilization from heat-related and other workplace injuries, (2) the effect of HB 2127’s repeal on the same outcomes, and (3) the mitigating effect of ordinances and their repeal on labor productivity in construction. The analysis is motivated by competing hypotheses about firms’ private incentives to protect workers from heat: if labor markets are competitive and risks well understood, ordinances should have little effect; if firms bear only a small share of heat-related costs or under-invest due to information frictions, municipal regulation could meaningfully improve worker health, and preemption laws like HB 2127 could be maladaptive. By providing credible causal evidence on the efficacy of local heat protection policy, this project aims to inform the evidence base for state and federal (e.g., OSHA) climate adaptation regulation.Competition for Space, Water, and Lithium Brine in Arkansas (with Beia Spiller)
Abstract
The Smackover Formation in southern Arkansas holds lithium dissolved in brine equal to an estimated 35 to 136 percent of previously documented U.S. lithium reserves, and development is expected to accelerate as policymakers move to secure domestic supply chains for critical minerals. Extraction there will rely on direct lithium extraction (DLE), a technology that has not been commercialized at scale and whose environmental impacts — particularly water consumption and contamination — remain poorly understood. We ask where that expansion would place lithium production in direct competition with drinking water supply and settled population. We combine machine-learning estimates of dissolved lithium concentration in Smackover brines with hydrologic watershed boundaries, U.S. Environmental Protection Agency records of drinking water utility intakes, and gridded population counts from ORNL LandScan Global 2024, intersecting them to rank watersheds jointly by resource potential and exposure. The resulting map identifies where abundant lithium coincides with drinking water infrastructure and concentrated population, and therefore where a contamination event or a water-use conflict would carry the largest consequences for human health and water security. The framework offers an early screening tool for siting decisions, regulatory oversight, and risk mitigation as domestic lithium production expands.Riders on the Storm: The Impact of Hurricanes and Tropical Storms on Platform Market Power in New York City’s Ride-Hail Market (with Xibo Wan)