Stephen Walt points us to this working paper (ungated) by Duke University’s Alexander Downes, who examines whether foreign imposed regime changes increase the likelihood of civil war. Downes finds that the answer is a clear “yes” if the purpose of the foreign intervention is to install a new leader in power as opposed to restoring a recently overthrown leader. The likelihood of civil war is especially strong in poor and/or ethnically heterogeneous countries and if the regime change occurs in conjunction with defeat in an interstate war. Downes uses matching to deal with the obvious problem that foreign countries are most likely to intervene in cases that are violent to begin with (matching tries to create balance between a “treatment” group and a “control” group based on observed characteristics such as geographic location, democracy, and development. Matching does not control for unobserved differences that may make civil war more likely in countries where foreigners intervene).
This seems like a very important and relevant study for thinking about Libya. The likelihood that there will be a civil war, at least in the short term, is high. It is not clear whether the intervention caused this, in the sense that the pre-intervention situation also closely resembled a civil war. It is likely that the intervention prolonged the civil war as it prevented a quick Gadhafi victory (see also Patrick Regan’s work on third party interventions and civil war). This type of nuance is difficult to capture in data analyses that are based on country-year data: if the intervention starts in the same year as the civil war, we do not really know what came first. Downes also estimates the long-term impact of foreign imposed regime change and finds that the interventions increase the likelihood of civil war long after their occurrence. These findings are probabilistic, so we can hope that Libya escapes this gloomy fate. Or we can argue that Libya has systematic characteristics that make it different from other interventions in ways that are not modeled in the paper. Nevertheless, we should take the possibility of a long civil war quite seriously.