UC Berkeley

Natural disasters and plagues have marked the course of human history, and remain among the greatest challenges and threats to economies and society. This dissertation uses a number of publicly-available datasets and natural experiments to explore how exposure to natural disasters affect households and societies in low- and middle-income African countries, with a particular focus on effects over time in agricultural communities. The disasters I study include desert locust swarms, floods, and the COVID-19 pandemic. In Chapter 1, I study the long-term effects of transitory agricultural shocks on the risk of violent conflict across Africa and the Arabian Peninsula using local variation in the areas exposed to desert locust swarms. In Chapter 2, I explore issues of measurement in determining what communities have been exposed to floods and how the use of different flooding definitions affects conclusions about the medium-term impacts of flood exposure on the incomes and activities of households in agricultural communities in Nigeria. In Chapter 3, I use a natural experiment during the COVID-19 pandemic to analyze how changes in household childcare burdens affect adults' labor supply in Kenya.

My first chapter studies how transitory agricultural shocks affect the local risk of violent conflict over time. I answer this question using data on conflict events desert locust swarms across 0.25 degree grid cells in Africa and the Arabian peninsula from 1997-2018. Using modern difference-in-differences and event study approaches, I compare areas exposed to a desert locusts swarm--a localized agricultural disaster--at some point during this period to nearby areas with similar underlying risk of exposure. Local variation in exposure is caused by swarm flight patterns along their migratory paths during major outbreaks, characterized by daily downwind flights of over 100km. I find that having been exposed to a locust swarm significantly increases the average annual probability of violent conflict in a cell by 0.8 percentage points (43%). This is a very large effect, equivalent to the effect of a 1.6 degree C higher temperature in the same year. Effects persist for at least 14 years and are driven by swarms arriving in crop cells during the main growing season.

I interpret the results using a model of occupational choice to explore income-related mechanisms. Persistent increases in conflict following a localized disaster suggest a decrease in the opportunity cost of fighting in affected areas. In line with this, I find that swarm exposure significantly reduces cereal yields in subsequent years, building on previous studies finding persistent effects of locust exposure on measures of household well-being and agricultural profits. Such effects indicate a permanent income mechanism for this severe transitory shock. This mechanism is not sufficient to explain the pattern of long-term impacts on violent conflict risk: the feasibility of engaging in violent conflict also matters. Increases in conflict risk are concentrated in years with active fighting groups in neighboring areas who may recruit or coerce individuals to join in violent conflict. Conflict therefore increases only when the reduced opportunity cost of fighting is combined with opportunities to fight. Patterns of long-term impacts on violent conflict are similar for severe droughts, indicating the mechanisms are not specific to locust shocks. Long-term impacts of transitory economic shocks on conflict risk add further motivation for policies mitigating the risk of such shocks and promoting household resilience and long-term recovery.

My second chapter explores how extreme weather events affect livelihood decisions of agricultural households over time, focusing on the devastating 2012 floods in Nigeria. I use nationally-representative panel household survey data together with satellite imagery to analyze how exposure to the 2012 floods affects household labor supply and income in subsequent years. I first show that identification of flooding exposure varies depending on whether survey reports or MODIS satellite imagery are used. Differences in characteristics of communities identified as flooded by only one definition imply specific measurement issues for each. These results indicate that more attention is needed to considering how flooding is defined and measured, and whether there are different impacts of exposure to different types of floods.

As in Chapter 1, I use a difference-in-differences design comparing changes in outcomes over time for households in communities exposed to flooding in 2012 against households in non-exposed communities that had a similar risk of flooding.I find no effects of flood exposure on engagement in wage employment or on total household income but a small significant increase in the probability of experiencing food insecurity using both definitions. The value of crop production falls, driven by a decrease in the value of commercial crops, while staple crop production either increases or remains stable. Using a MODIS-based flooding definition, I find that households exposed to floods significantly increase non-farm enterprise income relative to non-exposed households, as they reallocate labor from crop production to existing businesses. Using a survey-based definition, I find that flooding causes some farm households to become engaged in non-farm enterprise but without significantly increasing enterprise income while some non-farm households begin production of subsistence crops, with limited returns. The results emphasize that survey and satellite measures capture different types of flooding-related events which accordingly have different effects on exposed households. Studies of the impacts of floods should carefully consider their choice of flooding measure and the type of flooding they are most interested in analyzing.

My third chapter, coauthored with Dennis Egger and Utz Pape, identifies the impact of a shock to childcare and child labor on adult labor supply and intra-household allocation of productive activities in the context of COVID-19-related school closures in Kenya. Using nationally-representative bi-monthly panel data, we compare changes in labor supply after schools partially reopened in October 2020 for adults with children in a grade eligible to return against adults with children in adjacent grades. We find that a child returning to school increases adults' weekly work hours by 29% in the short run, concentrated among the most flexible margins of adjustment, particularly household agriculture.

Contrary to evidence from high-income settings, overall effects are not gendered. However, equal average labor supply responses for women and men are driven by different mechanisms particular to low- and middle-income settings. Women free up more time than men when childcare burdens fall but specialize more in childcare when returning students were net caregivers to younger siblings during school closures. Women also shoulder more of the reduction in child agricultural labor when students return to school, and shift from non-agricultural work into household agriculture (more easily combined with care of younger children). A back-of-the-envelope calculation suggests that school closures account for at least 40% of the overall drop in labor supply during the pandemic in Kenya, and a fall in GDP of 2.6%. Our results suggest policies increasing childcare access could substantially increase adult labor supply in low- and middle-income countries.

Climate change is increasing not only global temperatures but also the frequency and severity of extreme weather events. Globalization has made societies around the world more connected but also more vulnerable to the spread of pests and disease. Yet there is little evidence on how such disasters affect households in low-income countries over time, and how they affect labor and livelihood decisions of poor farm households in particular. The three chapters of my dissertation combine publicly-available data, economic theory, and rigorous identification based on modern econometric methods to contribute to policy discussions around mitigation and adaptation to climate change, structural transformation, and civil conflict, among the greatest concerns for policymakers in Africa and around the world.