Working Paper: CEPR ID: DP14300
Authors: Ashani Amarasinghe; Paul A. Raschky; Yves Zenou; Junjie Zhou
Abstract: We develop a network model of conflict in which players are involved in different battles. A negative shock in one locality affects the conflict in this locality but may also increase battles in path-connected localities depending on the location of the battle in the network and the strength of each locality involved in each battle. We then empirically test this model by analyzing the effect of local natural disasters on battles in Africa. We construct a novel panel-dataset that combines geo-referenced information about battle events and natural disasters at the monthly level for 5,944 districts in 53 African countries over the period from 1989 to 2015. At this fine temporal and spatial resolution, natural disasters are formidable exogenous shocks that affect the costs and benefits of fighting in a locality. We find that natural disasters decrease battle incidence in the affected locality and that this effect persists over time and space. This mitigating effect appears to be more pronounced in more developed localities. As highlighted by the model, these results can be explained by the fact that natural disasters divert fighting activity to surrounding localities, particularly those that are connected via geographic and road networks.
Keywords: Natural Disasters; Battle; Africa; Spatial Spillovers; Temporal Spillovers
JEL Codes: D85; O55
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| natural disasters (H84) | battle incidence (D74) |
| natural disasters (H84) | diversion of fighting activity to surrounding localities (H56) |
| battle incidence (D74) | diversion of fighting activity to surrounding localities (H56) |
| natural disasters (H84) | decrease battle incidence in the same district (H56) |
| natural disasters (H84) | increase likelihood of conflict spillover (D74) |