To determine the effect of shocks on civil conflict, it is critical to tailor the empirical approach to the persistence of shocks. I illustrate my point by revisiting a cornerstone of the literature on the economics of civil conflict, Miguel, Satyanath, and Sergenti’s (2004) study of rainfall and civil conflict in Sub-Saharan Africa. I find their approach to be inappropriate and their conclusions to be incorrect. For example, they conclude that higher rainfall levels are associated with significantly less civil conflict. I show that higher rainfall levels are actually associated with significantly more (not less) conflict in their data.
Data and estimation programs are available further below. For previous versions of the paper scroll down Papers, presentations, and data. Some results have changed because the most recent version of the paper uses the latest data.
The Economic Journal, May 2010. (This paper replaces Growth, Democracy, and Civil War.)
Transitory Economic Shocks and Civil Conflict
Miguel, Satyanath, and Sergenti (2004) argue that Sub-Saharan African civil conflicts 1981-1999 started following negative rainfall shocks. In Transitory Economic Shocks and Civil Conflict (May 2009) I use the same data and controls but conclude the opposite: conflicts started after positive rainfall shocks. Our conclusions disagree because I treat rainfall levels as mean reverting; MSS identify the effect of rainfall shocks on conflict outbreak assuming rainfall levels follow a random walk. The opposing conclusions highlight the importance of correctly capturing the persistence of economic shocks, an issue the conflict literature has not dealt with so far.
Recessions due to low rainfall triggered democratic change in Sub-Saharan Africa, which is consistent with the economic (window-of-opportunity) theory of political transitions. Our analysis quantifies democratic change by improvements in democracy scores as well as transitions from autocracy to democracy (an earlier (February 2008) version of the paper focused on the latter). The latest version eliminates a handful of inconsequential errors in the rainfall data that colleagues made us aware of.
I examine whether civil conflict is triggered by transitory negative economic shocks. My approach follows Miguel, Satyanath, and Sergenti’s 2004 paper “Economic Shocks and Civil Conflict” in using rainfall as an exogenous source of economic shocks in Sub-Saharan African countries. But I take into account that rainfall shocks are transitory (i.e. interannual rainfall growth is lowest following good years). Miguel, Satyanath, and Sergenti do not, and therefore conclude that conflict follows negative rainfall shocks when conflict in their data actually follows positive rainfall shocks.
Presentation at “Conflicts, Globalization, and Development” CEPR/PSE Workshop Paris, 13-14 November 2008 (the program of the workshop is available here). The presentation also drew on the work Markus Brückner and I are doing on commodity prices, growth, and civil war.
Agricultural risk and manufacturing earnings (presentation)
Slides (in pdf) of a at the LSE CEP/Economics Department Macro conference on February 8, 2008; maybe this additional brief explanation helps making clear what I am trying to get at
Growth, Democracy, and Civil War (with M. Brückner)