Resumen de Essays on war, forced displacement, hiv, and education
The field of Economics has benefited from the great improvements in data collection we have witnessed during the last decade. We now have within our reach datasets with high precision and frequency. The availability of higher frequency data allowed for a shift from cross-sectional data to panel data—same cross-section over time. Using panel data the researcher observes the same unit of analysis over time. The main advantage is that if researchers observe then same units over time, they can control for unobserved characteristics that while unknown, can be understood not to have much variation over time. In the recent years, data with precise geographical information is allowing reserachers to merge previously unrelated datasets using geographical location.
The data limitations of the past are being tackled by the development of methods addressing the unknown, and new data availability—and computing power—that allows us to enrich our data by merging it with data coming from all types of sources. The first chapter of this thesis uses panel data where the unit of analysis is a 2 mile by 2 mile square observed over two years to study forced displacement as a response to conflict in Iraq. Chapter 2 uses the exact geographical location of households in Rwanda, conflict during the 1994 genocide, and other geographical features to document the prevalence of wartime rape during the genocide in HIV levels a decade later. Chapter 3 uses a panel of schools in Spain to assess the impact of a bilingual education system on student learning outcomes.
Chapter 1 studies how internally displaced persons reacted as a response to violent conflict in Iraq during the war against ISIL between 2014 and 2017. I develop a network model that accommodates new data with exact geographical coordinates. The data on IDPs and conflict has a large spacetime variation that can be exploited by the estimation model. I contribute to fill the existing gap in the conflict literature regarding internally displaced persons by answering the following questions: How far from conflict do IDPs go? Where do IDPs shelter? How does conflict increase the probability of a location to host IDPs? How does conflict accumulate to trigger displacement? The highest concentration of IDPs is found within 2 miles of conflict and decreases with distance, disappearing beyond 40 miles. IDPs tend to cluster in highly populated areas, within 5 miles of a main road. Non-diverse ethno-religious areas host fewer IDPs relative to areas without a clear ethno-religious majority. An extra conflict event within 2 miles increases the probability of a grid cell to host IDPs by 30%. Forced displacement is triggered by conflict accumulating for two weeks at most.
Chapter 2 finds empirical evidence of wartime rape during the Rwandan genocide. I use HIV data 10 years after the genocide as a measure of the prevalence of rape to find that HIV levels in 2005 can be explained by the intensity of the genocide in the different Rwandan districts. The findings document both the prevalence of rape- usually stigmatized and hard to measure- and its dire lasting effects on the Rwandan population long after the crimes were committed. To establish causality, I exploit the exogenous variation in the accessibility to households during the genocide. I measure accessibility in terms of the distance from these households to the main roads, the rainfall over those roads during the genocide, and terrain ruggedness.
Chapter 3 analyses the impact of a bilingual education program aiming promote students' language proficiency and communicative competence in a language other than their own. Nowadays, bilingual programs are present worldwide responding to an increasing demand partially driven by the potential personal and economic benefits from being proficient in a foreign language. However, bilingual education increases the difficulty of learning academic content due to it being taught in a non-native language. To measure the importance of this effect, I utilize standardized test data and the Spanish-English bilingual program of Madrid. The findings show a small but significant negative impact of the program on the performance of students in English-taught content. The negative effect is stronger around the median of the student's ability distribution
