Do UN Interventions Cause Peace? Using Matching to Improve Causal Inference

Abstract

Previous statistical studies of the effects of UN peacekeeping have generally suggested that UN interventions have a positive effect on building a sustainable peace after civil war. Recent methodological developments have questioned this result because the cases in which the United Nations intervened were quite different from those in which they did not. Therefore the estimated causal effect may be due to the assumptions of the model that the researchers chose rather than to peacekeeping itself. The root of the problem is that UN missions are not randomly assigned. We argue that standard approaches for dealing with this problem (Heckman regression and instrumental variables) are invalid and impracticable in the context of UN peacekeeping and would lead to estimates of the effects of UN operations that are largely a result of the assumptions of the statistical model rather than the data. We correct for the effects of nonrandom assignment with matching techniques on a sample of UN interventions in post-Cold-War conflicts and find that UN interventions are indeed effective in the sample of post-civil-conflict interventions, but that UN interventions while civil wars are still ongoing have no causal effect.