Somalia is one of the poorest countries in Sub-Saharan Africa, with 69% of the population living under the standard international poverty line of US$1.90 in 2017-18. In 2004, an interim central state was established with the aim of bringing political stability across Somali regions.
The political transition culminated with the establishment of the Federal Government of Somalia in 2012 and a first electoral process in 2017. The elected government has aimed to improve national security conditions, yet the opportunity to ensure a development trajectory still faces many challenges, among them terrorist attacks (World Bank, 2018).
At a first glance, the consequences of a terrorist attack might seem small and contained given that they usually affect a small fraction of the population and the economy. Yet, several studies suggest sizable effects on economic outcomes (Abadie and Gardeazabal, 2008). Further, nearly two-thirds of the poor around the world are projected to live in conflict-affected countries by 2030, including Somalia. Therefore, it is important to shed light and improve our understanding on the links between conflict and poverty.
This paper estimates the immediate (within a week) impact of terrorist attacks from Al-Shabaab against civilians in Somalia using micro-data from two waves of the Somali High Frequency Survey (SHFS), combined with geo-tagged information on attacks. We exploit the spatial and time variation of interviews through a difference-in-difference identification strategy that compares outcomes of control and exposed households, before and after terrorist incidents. We also derive a shift-share instrument using changes in the number of US air/drone attacks against Al-Shabaab and employ an instrumental variables approach. We provide evidence to support the validity of our identification strategies and that our estimates are robust to different specifications, samples considered and several sensitivity checks.
Our results suggest that consumption of households exposed to terrorist incidents decreases by 33%, mainly driven by a decline in food consumption. The reduction in consumption increases poverty and the depth of poverty among the poor. The impact on consumption seems to be associated to a smaller share of household members (aged 15 to 50) working and earning income after an attack.
In addition, we document that the negative impact on consumption is clustered within a 4 kilometer radius from the incident and has a heterogeneous impact, not affecting households in the top 20% of the consumption distribution. The perception of police competence also worsens as a result of a terrorist incident. The literature models terrorists as rational actors, with terrorism having large consequences on economic outcomes, besides the loss of life, damage to persons and negative psychological effects. Conflict can also lead to sharp increases in poverty and vulnerability and other adverse outcomes.7 Our findings are in line with the disruption that could be expected from a terrorist attack. We contribute to the literature on the intersection between poverty and adverse shocks in developing countries, as well as to the policy debate by quantifying the impact of terrorist attacks on consumption and poverty, describing which households are affected by such incidents and the mechanisms through which this is likely to occur. Most of the empirical literature on the effects of terrorism on economic outcomes has relied on data aggregated at some geographical level (district, region or country), while the growing body of research exploiting micro-data to understand the effect of various shocks on poverty has not analyzed the effect of terrorism. To our knowledge, this is the first study to measure the causal impact of terrorism on consumption and poverty using household-level data in a fragile and conflict-affected country. The paper is structured as follows: The next section discusses the related literature on the effects of terrorism and multiple shocks on welfare conditions for households. Section 3 describes the data sources, sample considered and the definition of households exposed to terrorist incidents, besides specifying the identification strategies. Section 4 presents the results and extensions. Section 5 discusses multiple robustness checks and supplementary OLS estimates, while Section 6 contains our concluding remarks.