Published papers

“Machine learning predicts which rivers, streams, and wetlands the Clean Water Act regulates” (with S. Greenhill, S. Wang, D. Keiser, M. Girotto, J. Moore, N. Yamaguchi, A. Todeschini, and J. Shapiro) Science (2024). Available here.

We assess which waters the Clean Water Act protects and how Supreme Court and White House rules change this regulation. We train a deep learning model using aerial imagery and geophysical data to predict 150,000 jurisdictional determinations from the Army Corps of Engineers, each deciding regulation for one water resource. Under a 2006 Supreme Court ruling, the Clean Water Act protects two-thirds of US streams and more than half of wetlands; under a 2020 White House rule, it protects less than half of streams and a fourth of wetlands, implying deregulation of 690,000 stream miles, 35 million wetland acres, and 30% of waters around drinking-water sources. Our framework can support permitting, policy design, and use of machine learning in regulatory implementation problems.

“Accounting for ecosystem service values in climate policy” Nature Climate Change (2022). Available here.

Ecosystem services are often omitted from climate policies due to difficulties in estimating the economic value of climate-driven ecosystem changes. However, recent advancements in data and methods can help us overcome these challenges and move towards a more comprehensive accounting of climate impacts.

“Wetlands, Flooding, and the Clean Water Act” (with Charles A. Taylor). American Economic Review, 112.4 (2022): 1334-63. Available here.

In 2020 the EPA narrowed the definition of ‘Waters of the United States', significantly limiting wetland protection under the Clean Water Act. Current policy debates center on the uncertainty around wetland benefits. We estimate the value of wetlands for flood mitigation across the US using detailed flood claims and land use data. We find the average hectare of wetland lost between 2001 and 2016 cost society $1,840 annually, and over $8,000 in developed areas. We document significant spatial heterogeneity in wetland benefits, with implications for flood insurance policy and the 50% of ‘isolated’ wetlands at risk of losing federal protection.

The effect of large-scale anti-contagion policies on the COVID-19 pandemic (with S. Hsiang, D. Allen, S. Annan-Phan, K. Bell, I. Bolliger, T. Chong, L. Huang, A. Hultgren, E. Krasovich, P. Lau, J. Lee, E. Rolf, J. Tseng, and T. Wu). Nature 548, 262-267 (2020). Publication. Replication data and code.

Governments around the world are responding to the coronavirus disease 2019 (COVID-19) pandemic with unprecedented policies designed to slow the growth rate of infections. Many policies, such as closing schools and restricting populations to their homes, impose large and visible costs on society; however, their benefits cannot be directly observed and are currently understood only through process-based simulations. Here we compile data on 1,700 local, regional and national non-pharmaceutical interventions that were deployed in the ongoing pandemic across localities in China, South Korea, Italy, Iran, France and the United States. We then apply reduced-form econometric methods, commonly used to measure the effect of policies on economic growth to empirically evaluate the effect that these anti-contagion policies have had on the growth rate of infections.

Working papers

Can Removing Development Subsidies promote adaptation? The Coastal Barrier Resources System as a Natural Experiment (with Penny Liao, Sophie Pesek, Shan Zhang, and Margaret Walls)

As natural disasters grow in frequency and intensity under climate change, limiting populations and properties in harm's way will be one important facet of adaptation. This paper examines the Coastal Barrier Resources Act of 1982, which eliminated federal incentives for development in designated areas along the Atlantic and Gulf coasts known as the Coastal Barrier Resources System (CBRS). We introduce a new research design to estimate the causal effect of the policy that identifies plausible counterfactual areas using machine learning and matching techniques. We find that CBRS designations lower development density by 85% inside the designated areas, but increase development in neighboring areas by 20%. We also present new evidence on flood protection benefits, property values, and changes in demographic characteristics in the affected areas. Our results inform ongoing debates regarding cost-effective policy options for discouraging over-development in areas at risk of climate change.

Global High-Resolution Estimates of the United Nations Human Development Index Using Satellite Imagery and Machine Learning ( with L. Sherman, J. Proctor, H. Tapia, and S. Hsiang). NBER Working Paper. Replication Code. Data.

The United Nations Human Development Index (HDI) is arguably the most widely used alternative to gross domestic product for measuring national development. This is in large part due to its multidimensional nature, as it incorporates not only income, but also education and health. However, the low country-level resolution of the global HDI data released by the Human Development Report Office of the United Nations Development Programme (N=191 countries) has limited its use at the local level. Recent efforts used labor-intensive survey data to produce HDI estimates for first-level administrative units (e.g., states/provinces). Here, we build on recent advances in machine learning and satellite imagery to develop the first global estimates of HDI for second-level administrative units (e.g., municipalities/counties, N = 61,591) and for a global 0.1 × 0.1 degree grid (N=806,361). To accomplish this we develop and validate a generalizable downscaling technique based on satellite i! magery that allows for training and prediction with observations of arbitrary shape and size. This enables us to train a model using provincial administrative data and generate HDI estimates at the municipality and grid levels. Our results indicate that more than half of the global population was previously assigned to the incorrect HDI quintile within each country, due to aggregation bias resulting from lower resolution estimates. We also illustrate how these data can improve decision-making. We make these high resolution HDI estimates publicly available in the hope that they increase understanding of human wellbeing globally and improve the effectiveness of policies supporting sustainable development. We also make available the satellite features and software necessary to increase the spatial resolution of any other global-scale administrative data that is detectable via imagery.

Accounting for Unobservable Heterogeneity in Cross Section using Spatial First Differences (with Solomon Hsiang) NBER WP 25177. Working Paper. Replication data and code.

We develop a cross-sectional research design to identify causal effects in the presence of unobservable heterogeneity without instruments. When units are dense in physical space, it may be sufficient to regress the “spatial first differences” (SFD) of the outcome on the treatment and omit all covariates. The identifying assumptions of SFD are similar in mathematical structure and plausibility to other quasi-experimental designs. We demonstrate the SFD approach by recovering new cross-sectional estimates for the effects of time-invariant geographic factors, soil and climate, on long-run average crop productivities across US counties — relationships that are notoriously confounded by unobservables but crucial for guiding economic decisions, such as land management and climate policy.

Estimating an Economic and Social Value of Forests: Evidence from Tree Mortality in the American West. Available here.

Linkages between healthy forests and human well-being are often theorized, yet the magnitude of benefits remains unknown. This paper uses a natural experiment to assess the welfare consequences of changes in forest health across the American West. My empirical analysis relies on plausibly random variation in tree mortality generated by the thermal threshold at which cold-induced mortality occurs in bark beetles. I find that forest die-off has significant and economically meaningful impacts on both the market value of forests and the non-market benefits these ecosystems provide. I estimate that over the last two decades, tree mortality in the American West decreased the value of timber tracts by $1.1 billion, decreased home values by $16.5 billion, and increased damages from air pollution, wildfire, and floods by a combined $921 million. In a back-of-the-envelope calculation, I find that the death of a tree in my sample costs society $43 in foregone benefits

Using Historical Aerial Photography and Machine Learning to Map 20th Century Global Change ( with S. Hsiang, A. Madestam, and A. Tompsett)

Satellite imagery has transformed our understanding of both the natural world and human well-being, but the instrumental record only begins in 1972. We develop an approach to extend these data backwards in time an additional 30 years. We digitize the only complete archive of 1.6 million aerial photographs collected during the processes of mapping what was then the British Empire. This archive surveys more than 60 former colonies, where granular data from other sources is notoriously sparse, repeatedly from the 1940s to the 1990s. We develop a novel machine learning approach to assemble these photographs into mosaics comparable to modern satellite observations. We then generate large-scale, gridded datasets that provide measures of population density, land use, and infrastructure at 1km resolution.

Opportunities for Increasing the Environmental Justice Impact of Earth Observations. Available here.

Environmental justice is an important social priority and has become a policy goal at the federal level. Executive Order 14008, Tackling the Climate Crisis at Home and Abroad, requires all federal agencies to develop programs, policies, and activities to address the disproportionately high adverse environmental impacts faced by marginalized communities. This raises the question of how NASA can increase the impact of its scientific outputs along EJ dimensions. How can the agency’s existing data products be leveraged to enable progress on EJ-related questions? What new scientific information can NASA produce to help reduce environmental inequities? Produced as part of the VALUABLES Consortium, this paper aims to outline ways that NASA can use its data products to promote EJ and overcome the barriers faced in the use of satellite data to influence decisionmaking.

Research in progress

“Costs of Land Use Regulation under the Clean Water Act” (with Joseph Shapiro and Charles Taylor)

“Willingness to pay for flood-resistant homes” (with Jaimie Choi, Billy Babis, and Michael A Toman)