The empirical studies that Erica Chenoweth has highlighted in the past few days give ample reason for pause about the likelihood that a U.S. led military intervention will reduce civilian killings or end the war in Syria. That said, I am somewhat skeptical that we can generalize from these studies to the case of Syria.
All of these studies investigate whether the likelihood of some outcome (high civilian killings, ending a civil war, etcetera) becomes more likely when an intervention has taken place than when it has not. The answer for the most part is that on average few good things happen post military intervention. There is the obvious problem that one can only uncover correlations this way. For instance, it may be that interventions only take place in the worst conflicts.
There is an even more basic problem. As Jon Western points out, there is tremendous variation in the scope, size, and purpose of interventions. Conflict situations similarly differ enormously. The correlations are thus average associations between a very heterogeneous “treatment” (interventions) among very heterogeneous units of analysis (cases of civil war/conflict). Think of it as trying to estimate what effect medicine has on your health when we group together patients with different diseases and thus different medicines. We may find a positive effect of “medicine” but we don’t know if this was because all medicines on average improve health or if there are some that work exceptionally well while others leave patients worse off.
This doesn’t mean that these studies aren’t good or important. At times, with important questions like these, you have to do the best you can with what little you have. It is also important, as Erica did, to bring the findings of these studies into the public debate. It does mean that we should exercise some caution in interpreting their policy implications. Erica’s post on what types of interventions are likely to work is exceptionally useful in that regard. Yet, that post highlights very different types of interventions than what we are likely to see in Syria. Multilateral peacekeeping operations are more likely to help end civil wars than other types of interventions. Yet, I am not sure whether we should conclude from the existing research that a Syrian intervention is unlikely to work or that we haven’t seen enough comparable cases to draw reliable conclusions?
Syria is unusual in several respects but probably its most important distinction is that it is an episode of “mass killing” even before the intervention takes place. How common is this in the historical record? Benjamin Valentino has carefully documented episodes of mass killings. The modal episode of mass killings had no meaningful intervention during the worst mass killings, such as the Armenian genocide in Turkey or the genocide in Burundi. His list contains several examples of cases where outside intervention may have induced mass killings; Vietnam and Iraq may be the best examples. There are only a few episodes where the intervention came after the mass killing had already started. The interpretation of the effect of interventions is debatable in all these cases. Vietnam removed the Khmer Rouge regime in Cambodia which ended mass killings although it is questionable as a case of humanitarian interventions. Bosnia and Rwanda had small interventions that may have been counterproductive and may have led to increased killings. In both cases larger intervention forces eventually helped stabilize the countries.
I cannot claim to be an expert on all these episodes so I am not going to make a very precise claim here but it seems to me that there are not very many examples where a mass killing was already underway and an outside intervention took place and many fewer where these interventions obviously worsened the carnage.
The type of intervention may also be important. If we believe the Administration, the goal is to punish the Syrian military and government for their use of chemical weapons. Despite the articles written over the last few days arguing that a limited war is very difficult, I have little doubt that even a three-day bombing campaign by the U.S. military could do a lot of damage to the offensive capabilities of the Syrian army. This may or may not reduce civilian casualties but it certainly is very different from the interventions in Iraq, Rwanda or Bosnia. This puts both an upper and a lower limit on the likely effect of an intervention: it is unlikely to lead to a stable peace but it may also not induce the kind of increases in violence one may see with occupation. (Of course it is always possible that the intervention will lead to more than I assume here.)
My main point is that I cannot think of many or even any comparable cases to Syria: where the intervention is a limited bombing campaign that takes place after the mass killing has long been under way. Kosovo could be an example or maybe Libya, but that would not have been in the datasets researchers have used given usual time lags. Most cases in the data on which the findings on the negative or limited impact of interventions were based were very different types of interventions in very different types of conflicts.
Could we still generalize from these findings to the Syria case? People may differ on this but I am skeptical. The literature on peacekeeping in civil wars has important implications for peacekeeping in civil wars but I am not so sure it can be generalized beyond that. This is clearly not a case of peacekeeping or even peacemaking or peace-enforcing. This does not at all mean that I am optimistic about the effects of an intervention. All I am saying is that history does not provide us with very reliable guidance on the matter.