Working papers

Surge of Inequality: How Different Neighborhoods React to Flooding

Recovery scenarios after flooding vary by locality, from permanent declines in economic activity to capital gains. This paper shows that divergent post-flood changes at the neighborhood level increased preexisting spatial polarization along property value, racial, and income lines. Using evidence from property sales in four US states affected by Superstorm Sandy in 2012, combined with buyers' demographics, I find that flooded properties in neighborhoods with high preexisting income had more high-income white buyers and higher sale prices than comparable non-flooded coastal properties, seemingly capitalizing on the flood and offsetting average drops. Using machine learning algorithms, I corroborate that of a rich set of preexisting place characteristics, neighborhood income best discriminates between most positively and most negatively affected properties. This evidence is consistent with a model of neighborhood segregation in which residential sorting—induced by credit-constrained households deriving higher disutility from flooding—rationally results in more high-income residents and higher property prices in initially higher-income neighborhoods. As coastal flooding is forecasted to increase, these results improve our understanding of the heterogenous impacts of floods, and on the existence of adaptive behavior, or lack thereof, after flooding.

Flood in the desert: health impacts of flood variability in the inner Niger delta, Mali (with Anna Tompsett)

The vast majority of climate impact studies focus on the relationships between local climate variables- such as temperature, precipitation, or hurricanes- and local outcomes. But climate change impact projections may systematically underestimate total impacts if they do not account for the ways in which local outcomes depend on more distant climate events, for example through upstream rainfall in major river basins. In the Niger Inland Delta, in Mali, agriculture depends almost entirely on the availability of surface water through an annual flood following the rain season in the Guinean highlands, located hundreds of kilometers away. We construct a temporal series of flood extent using remotely sensed imagery, combined with household survey data on the timing and location of births and infant deaths across Mali. Exploiting quasi-random spatial and temporal variation in annual flood extent in a differences-in-differences framework, we find that a 2% increase in local flood extent decreases infant mortality by 1.5 percentage points. Our results highlight the importance of accounting for the impacts of a changing climate that take place over large spatial scales.

Flood resilience in slums: community-responsive adaptation in Kibera, Nairobi (with J. Benard, V. Bukachi, J. Mulligan, L. Olang, and A. Tompsett)

[In the field]

A large and growing number of the world's population live in slums, where the twin trajectories of rapid urbanization and increased flooding driven by climate change collide. However, there is much we do not know about how flooding affects the lives of residents in informal settlements because high quality survey data from these contexts remains rare. Do rents price in flood risk? How well-informed are households about flood risk? Does flood risk disproportionately affect new arrivals who may be less well-informed about risk? What is the extent and what are the consequences of flood-induced displacement? In this study, we make progress on answering these questions by tracking 1500 slum residences through repeated surveys for up to four years and, exploiting variation in flood exposure, estimating the causal impact of flooding on health, property, rents, and displacement.

Effects of land tenure formalization on vegetation content in Guatemala

Land tenure insecurity, prevalent in many low income countries, has been hypothesized to lower investment in environmental services such as tree planting or soil erosion control. I test this hypothesis in the context of a land tenure formalization program in Guatemala, by exploiting temporal and spatial differences in the program rollout as a source of exogenous variation in tenure security. I combine administrative data on land tenure formalization with measures of vegetation content constructed using remote sensing imagery. Results show that even if vegetation content continued to decrease on average after tenure formalization following pre-trends, the speed of vegetation loss slowed down with the program. Different land covers followed distinct trends, with vegetation content in shrublands and grasslands significantly increasing with improved land tenure security. There is no evidence that areas more prone to soil erosion experienced differential changes in vegetation content. Overall, these results contribute to our knowledge on the impact of land tenure security, or lack thereof, on investment in environmental services in low income countries.