Praying for Rain? Water Scarcity and the Duration and Outcomes of Civil Wars

Abstract

Recent anecdotal evidence from the civil wars in Somalia and Yemen suggest that water scarcity may shape the dynamics of civil wars. While a considerable body of research has examined the connection between water scarcity (such as low rainfall) and the onset of civil war, very little research has examined how water scarcity may shape the duration and outcomes of civil wars. Looking specifically at rainfall, this paper argues that changes in access to water play a key role in the duration of civil wars. As rainfall declines, there is a reduction in resources available to both the government and the rebel group, leading to a stalemate in fighting. Furthermore, this paper argues that declines in rainfall are felt more acutely by rebel groups who seek to challenge the government through conventional warfare. This paper tests these propositions using hazard models. The results provide robust support for the propositions.

Keywords:

• Civil war duration
• water scarcity
• rebel strength
• resources

Notes

1. Huwais, Muhammad. ‘Yemen is Tearing Itself Apart over Water.’ News Week 30th January 2015.

2. ‘Water Wars’ Center for Investigative Journalism, 14th January 2017. https://www.revealnews.org/episodes/water-wars/

3. Young (1997) notes that they went as far as to paint sticks black so as to give the appearance that all of their soldiers were armed.

4. As a robustness check, this paper also uses the coding procedures employed by Greig, David Mason, and Hamner (2016) that identify rebel favorable vs. government favorable outcomes. Rebel favorable outcomes are civil wars that end in rebel victory, peace agreement and a ceasefire. Government favorable outcomes are coded as all conflicts that end in either a government victory or with low activity. The conclusion of these results were roughly the same as the analysis presented in the results section: rainfall, in conjunction with rebel strength, increases the risk of civil wars ending in rebel favorable outcomes.

5. Salehyan and Hendrix (2014) note that the 10-year moving average is preferable to the total panel mean as there is significant trending over time for climate variables.

6. To account for unobserved country level variance, models were also estimated with First Differences of rainfall.

7. As a robustness check, I included a control for the percentage of the national economy that is drawn from the agricultural sector. The results remain largely unchanged.

8. For ease of interpretation, rebel group strength is measured through a series of binary variables as opposed to an ordinal scale. The results of the analysis can be found in Table A2 in the Appendix Section.