Border Fences and the Mexican Drug War
Border security is rapidly expanding globally as states build fences and walls to deter migration and smuggling. How do these barriers affect the behavior of the criminal organizations that engage in illegal cross-border activities? I provide evidence from the US-Mexico border, where 649 miles of fencing were built between 2007 and 2011. I use novel fine-grained data on violence from death certificates, engineering maps of the border fence, and a difference in differences research design to show that construction of the fence did not change violence in Mexican localities near the border fence, but caused at least 2,000 additional deaths in localities that provided access to alternate smuggling routes into the US. I hypothesize that by creating a shock to the value of territory near the border controlled by drug cartels, construction of the border fence undermined tacit agreements about territorial control that previously reduced violent conflict. This caused competition between cartels over territory providing access to newly more valuable alternate smuggling routes. This highlights the conditions under which non-state actors violently compete to control territory and provides stark evidence of the international effects of domestic policies.
Information Cascades and Refugee Crises [Under review]
Refugee crises repeatedly surprise the international community with their size and suddenness, yet we know little about what drives them. I develop a theory of refugee crises in which civilians living in conflict zones make individual decisions to flee in response to new information about the risk of victimization in war. The information conveyed by observing refugees fleeing can result in an information cascade, in which waves of refugees fleeing cause other civilians to increase their beliefs about the risk, increasing the numbers of subsequent refugees. To test this theory, I construct a geocoded village-day level dataset of refugee flows, violence against civilians, and the actions of armed groups during the Kosovo war. I develop an instrumental variables estimation strategy using the spatial network of villages connected by roads and the fact that refugees fled toward a single border crossing to estimate the causal spillover effect of refugees fleeing. I find that on average a refugee fleeing causes more than one additional civilian to flee.
Cell Phones, Rumors, and Internal Displacement in Civil War [Under review]
Cell phones have become ubiquitous in conflict zones. However, little is known about the effect of communication technology on migration during civil war. I hypothesize that access to cell phone networks causes people to flee by facilitating the rapid spread of rumors of violence within insecure populations. This may increase the impact of violence on civilian displacement as uncertain but plausible information about violence travels through social networks. To identify the effect of cell phone coverage on displacement during conflict I combine maps of cell phone coverage in Colombia with detailed data on insurgent violence against civilians and internal displacement and employ a spatial regression discontinuity research design along the cell phone coverage boundary. I find that access to cell phone networks causes an average of over 100 more people to flee from the hundreds of municipalities with cell phone coverage. In support of rumors as a mechanism, I show that this effect is not driven by changes in insurgent violence, which is unaffected by network coverage.
Technology and the Formation of National Identity: Evidence from Africa
with Donhyun Danny Choi and Anna Schultz
An important part of state-building is the creation of a national identity. A long line of theoretical literature leads us to expect that technological advancements facilitate the adoption of a national identity. We investigate the effect of one dimension of technological progress: the expansion of mobile phone coverage across sub-Saharan Africa, which is believed to induce broad changes in political and social life. We use a novel combination of geocoded public opinion data and fine-grained data on mobile phone coverage boundaries. Applying a geographic regression discontinuity design, we show that access to mobile technology decreases the likelihood that an individual identifies with the nation by around 5–7 percentage points. We hypothesize that the decrease in national identification is a result of the ethnically-polarized nature of political rhetoric that is often shared via cell phone networks, especially in the run-up to elections. To establish support for this mechanism, we exploit as-if random variation in the timing of individuals’ survey interviews to presidential elections. Our analysis suggests that the proximity to elections intensifies the effect of mobile coverage.