Working Papers
Flooding and Health Outcomes in Zambia
(Job Market Paper Draft Coming soon. Slides available upon request)
Floods are becoming a common phenomenon in African countries, and Zambia, a landlocked country in southern Africa, is no exception. Floods have been linked to the spread of vector-borne diseases, such as malaria, and waterborne diseases, including cholera and diarrhea. In Zambia, malaria is still the leading cause of morbidity, and diarrhea is among the top ten causes of morbidity. According to the National Malaria Elimination Centre (by the Ministry of Health), annual malaria expenditures in Zambia for the year 2021 stood at $61 million, with 85% of the funding outside of this value coming from external donors. The most recent 2023 cholera outbreak in Zambia required approximately $5 million to combat it, and waterborne diseases in general cost the country around 1.3% of its annual GDP.
In this paper, I ask the research question: by how much does flooding increase the incidence of malaria, and waterborne diseases such as cholera, typhoid, dysentery, and non-typhoid non-dysentery diarrhea? To answer the question, I use satellite 2-day flood data from NASA MODIS, and monthly health facility visits on confirmed (tested) cases of malaria and cholera, and clinical (based on symptoms) cases of typhoid, dysentery, and non-typhoid non-dysentery diarrhea from the Ministry of Health in Zambia for the years 2012 to 2024. I analyze the data at the ward level, which is the smallest administrative boundary in Zambia. I use a fixed effects approach, adding month-year fixed effects to capture seasonality, ward fixed effects to control for baseline infrastructure, hygiene norms, and other ward-specific time-invariant factors, and constituency by year fixed effects to control for constituency-specific year-by-year changes, such as the utilization of the constituency development funds. I then control for geographic and economic indicators, as well as the population and number of health facilities in each ward.
The main result of the analysis in this paper using pseudo maximum likelihood with high-dimensional fixed effects (ppmlhdfe) estimation is that for the entire population, floods increase the incidence of malaria and cholera. When a ward floods, the number of malaria cases per 100000 of the population will increase compared to when the ward does not flood, and this effect is statistically significant. One more flood will also increase the number of malaria cases per 100000 of the population. For cholera, when a ward floods, the number of cholera cases per 100000 of the population increases than when the ward is not flooded, and this effect is statistically significant. The number of floods does not matter in the increase in the incidence of cholera. Floods, however, do not affect the incidence of typhoid, and for dysentery and diarrhea, floods have a mitigating effect in that they reduce the incidence of these two diseases.
Work in Progress
Development and Flooding in Zambia
I study how rapid development, as measured by deforestation and the appearance of impervious surfaces in Zambia, affects the prevalence of floods from 2012 to 2019. I use deforestation data from Global Forest Change (Hansen et al 2013), and dynamic impervious surfaces data (Huang et al 2021) to examine the impact, leveraging the year-to-year changes in deforestation and impervious surfaces as a source of variation. Preliminary results show that deforestation has a larger effect on increasing floods than increasing total impervious surface area or even increasing marginal impervious surface area.
Automatic Grade Progression, Education Attainment, and Child Labor
I examine the impact of the automatic grade progression policy implemented by the Zambian government from 2013 to 2022 on educational attainment and child labor for affected individuals of school-going age. I use the annual 2012 to 2022 Labor Force Participation Surveys compiled by the Zambia Statistics Office (Zamstats), and school-level scores on the national grade 7 examinations to assign treatment and control groups based on district-level failure rates. An individual of grade 7 school-going age residing in a district with a high level of failure rate (above 50% of total is assigned as being treated, and individuals with similar characteristics (as the treated cohort) living in districts with low failure rates serve as controls. I use a difference-in-difference framework, with 2018, the year in which the policy was emphasized (took effect), as the cutoff for a pre and post-automatic grade progression. The outcomes of interest are the years of schooling and the number of hours spent working for money for children under the age of 18.
