Gone with the Wind: Measuring the effect of desert locust outbreaks

Authors: Myriam Marending and Stefano Tripodi

Abstract: This paper examines a measurement to disentangle the impact of geography in weather-related events. Using a desert locust outbreak in Ethiopia, we simulate counterfactual shock exposure of villages to measure the causal effect of locust swarms on farm profits. Our identification strategy relies on modelling swarms’ movements based on local climatic and environmental characteristics. We compare areas identified as treated to their control counterparts, and propose to condition on their expected exposure to correct for the selection bias. The findings suggest a sizeable farm profit loss in areas hit by the shock, and a sizeable non random exposure bias. [SSRN Working paper]


This figure shows the location of the Ethiopian Socio Economic Survey enumeration areas. This figure shows the geographical distribution of treated (triangle) and control (circle) enumeration areas of the Ethiopian Socio Economic Survey in 2014. The treatment is assigned using the NOAA-HYSPLIT model to calculate three days flight routes, exploiting the passive flier of the insects. This allows to simulate counterfactual shock assignments using historical climatic and environmental data holding fixed the geography that governs the shock exposure. The figure illustrates that differential geography causes systematically different expected locust exposures across locations. We address this potential source of bias by controlling for expected exposure to desert locust in our regressions.