Last updated on April 6, 2014
The abstract of a great new article by Jean-François Maystadt and Olivier Ecker in the American Journal of Agricultural Economics titled “Extreme Weather and Civil War: Does Drought Fuel Conflict in Somalia through Livestock Prices?”:
A growing body of evidence shows a causal relationship between extreme weather events and civil conflict incidence at the global level. We find that this causality is also valid for droughts and local violent conflicts in a within-country setting over a short time frame in the case of Somalia. We estimate that a one standard deviation increase in drought intensity and length raises the likelihood of conflict by 62%. We also find that drought affects conflict through livestock price changes, establishing livestock markets as the primary channel of transmission in Somalia.
The emphasis is mine. And in case anyone wondered, I was not a reviewer for this paper (and generally, I try not blog about the papers I get to review…)
Marc – are these large effects? From glancing at the paper: confusingly the authors use a count variable of conflicts, but then switch to talking about the probability of conflict. Glancing at the standard deviation on the conflict variable, I’m wondering if drought is really one of the main drivers here….
Thanks a lot for your interest in the paper. Let me reply in two steps:
1) You are right that we have used a linear model despite the fact our dependent variable is a count variable (note that we obtain similar results when we transform our dependent variable with a dummy variable equal to one if there is at least one conflict event). A linear specification of a two-stage least-squares, fixed-effect (2SLS-FE) model is well-established in the conflict literature. Non-linear models may yield biased estimates in a panel data setting with relatively rare events. We mention that around footnote 8.
2) We compute the size of the effects in percent by expressing the partial effect (the within-region standard deviation*coefficient) divided by the mean value of conflict (Of course, we are only looking at within-region variation). Note that for comparability reasons, we follow an approach that is very consistent with the results reported by Hsiang, S., M. Burke, and E. Miguel. 2013 (Quantifying the Influence of Climate on Human Conflict. Science 342 :6151). Our results stand at the upper part of the distribution of conflict risk impacts reviewed by Hsiang, Burke, and Miguel (2013).This may be explained by the exceptionally high vulnerability of pastoralist livelihoods in Somalia and in large parts of the Horn of Africa. We are also working on such an hypothesis for Sudan (http://www.ifpri.org/publication/local-warming-and-violent-conflict-north-and-south-sudan ). We are working on a new draft for this paper. That working hypothesis will be further discussed at the 2020 IFPRI conference (we are working on a policy brief on the issue) with policy-makers and practitioners: http://www.2020resilience.ifpri.info/
Best, JF
Thank you for this fruitful exchange, Matt and JF. And JF, congrats again on a very nice paper.