Working Papers
Religious Institutions and Environmental Regulation — Job Market Paper
Abstract (click to expand): Land-clearing fires in Indonesia caused an estimated 100,300 excess deaths in 2015, and state-sponsored bans have repeatedly failed because the state can detect fires but cannot attribute them to the farmers who set them. In 2016, Indonesia's highest Islamic authority declared environmentally harmful burning religiously forbidden, a ruling that carried no legal penalty and reached roughly 300,000 mosques through Friday sermons. Comparing adjacent villages across Kalimantan's centuries-old religious frontier, I find that fire detections in Muslim-majority villages fell sharply relative to paired neighbors after the ruling, and the estimated decline persisted through 2024. To my knowledge, this paper gives the first evidence that a religious prohibition can reduce a large environmental harm where state enforcement has not.
Legalizing Fire on Private Land Increased Burning and Forest Loss on Public and Indigenous Lands
(with Kathryn Baragwanath, Cesar B. Martinez-Alvarez, and Flora (Yifan) He)
Submitted to PNAS
Abstract (click to expand): Nearly three-quarters of the world's forests sit on government-administered land, where property rights are often weak and fire is the cheapest way to clear vegetation and claim territory. Governments across the tropics nonetheless face recurring pressure to relax restrictions on burning, framed as pro-poor rural development, with little evidence on the consequences for surrounding public lands. We study Bolivia's Supreme Decree 3973 of July 2019, which authorized land clearing and controlled burning on private and communal lands in the departments of Santa Cruz and Beni. Applying a doubly robust difference-in-differences design to satellite records of fire, burned area, land cover, and land tenure from 2015 to 2022, we find that the decree raised fire incidence by 152 percent. Fires doubled on the communal and large private lands the decree authorized, but rose furthest where burning remained illegal: tripling on untitled public land and protected areas and rising more than fivefold on Indigenous territories. Small private properties, the decree's stated beneficiaries, showed no significant change. The additional fires burned predominantly forests and natural grasslands rather than existing farmland; untitled public land suffered the largest primary forest losses of any tenure type and a 13 percent expansion of agricultural land. These fires were ignited on public land, not spread from private properties, while formal sanctions for illegal burning fell from 332 to 39. Fire deregulation on private land thus operates as a signal of impunity that licenses encroachment by fire on public natural landscapes, a spillover that assessments of frontier fire policy must account for.
Strategic Adaptation & Satellite Blind-Spots: Evidence from Northwestern India and Pakistan
(with Kathryn Baragwanath)
Abstract (click to expand): India enforces its crop-residue burning bans with detections from a polar-orbiting satellite that passes at a fixed afternoon hour. Using a geostationary satellite that images northwestern India every ten minutes, we observe the fires the enforcement system misses and find that roughly three quarters of fire activity now occurs outside the monitoring window. Official counts fell by roughly a third while burning continued in the evenings, consistent with farmers shifting when they burn rather than whether they burn in response to predictable enforcement.
Access to Justice in Times of Pollution: Evidence from South Asia
(with Ali Bakhtawar)
Abstract (click to expand): We ask whether exposure to crop-residue smoke changes the decisions of India's district courts. We link daily records from millions of cases to nearby fire detections and wind, summing each court's downwind exposure over the preceding days so that variation comes from shifts in the wind rather than from where courts or fires sit. Judges in downwind courts acquit a smaller share of defendants and decide fewer cases on the day of filing, while bail decisions and caseloads do not change, consistent with smoke changing how and when judges decide rather than how much work they finish.
Work in Progress
Re-optimizing Fires: Increasing Information for Crop Residue Fires in Pakistan
(with Kathryn Baragwanath, A. Patrick Behrer, Jinjing Chen, and Caleb Robinson)
Abstract (click to expand): With Microsoft's AI for Good Lab, we build a deep-learning system that detects crop-residue fires field by field in 3-meter satellite imagery for the Government of Punjab, Pakistan. We validated the system against 1,270 ground-truthed fields, and it now feeds a monitoring portal at the provincial Environmental Protection Department. We trace each fire's smoke on the wind of the day it burned to measure the harm farmers would avoid by burning on a different day, and we are designing a field experiment comparing burn-day advisories with payments to test which instrument reduces fire-driven pollution most.
Distributional Effects of Burning: Evidence from Bolivia
(with Kathryn Baragwanath and Flora (Yifan) He)
Abstract (click to expand): We apply the HYSPLIT dispersion model to trace smoke from fires detected in the years around Bolivia's 2019 land-clearing decree and link the resulting exposure to municipal records of hospitalizations for respiratory and cardiovascular illness. We measure the health costs the decree's fires imposed on downwind communities, so that the decree's full cost includes the health lost downwind as well as the forest its fires burned.
Land Use and Technological Development in Pakistan
(with Klaus Deininger)
Publications
(with Husnain Fateh, Matthew Gibson, Arman Rezaee, and Sanval Nasim)
The Economic Journal, August 2025
Abstract (click to expand): Scarce information and human capital may make it difficult to form accurate expectations, limiting responses to uncertain environmental threats like air pollution. We study two cross-randomised interventions in Lahore, Pakistan: (i) general training in forecasting and (ii) provision of air pollution forecasts. Both reduced the subjects’ own air pollution forecast errors; the training effect suggests that modest educational interventions can durably improve forecasting skills. Forecast receipt increased demand for protective masks and the responsiveness of outdoor time to pollution. Forecast recipients were willing to pay 41% of their total mobile phone costs for continued access, consistent with welfare gains from adaptation.
(with Blair, G.; Christia, F.; Weinstein, J.; and others)
Cambridge University Press, September 2024
(with Graeme Blair, Fotini Christia, Jeremy Weinstein, and others)
Science, 2021
(with Husnain Fateh, Hassan Khan, Mashood Nasir, and Nauman Zaffar)
Energy Policy, September 2018