Though he has stated his preference between types of strikes, Mohmand is still begging the question of whether airstrikes of any kind serve to reduce militant violence. I thought this would be a good opportunity to review the two best research papers I've seen on this question as it applies to Afghanistan and Pakistan. Both papers are meticulous yet they seem to disagree on the efficacy of airstrikes for reasons we don't yet understand.
Johnston and Sarbahi on drone strikes along the border
In the latest version of "The Impact of U.S. Drone Strikes on Terrorism in Pakistan and Afghanistan," Patrick Johnston and Anoop Sarbahi examine data from 2004 to 2011 to gauge whether or not drone strikes affect the frequency or lethality of terrorist attacks. Their geographical focus is on northwest Pakistan and the bordering regions of Afghanistan. Johnston and Sarbahi test a number of hypotheses, but the key findings in the paper are that drone strikes are associated with reductions in terrorist violence and that there is no spillover of violence into nearby areas. The authors make sure to caveat their results, noting that they cannot possibly measure recruitment effects and that the paper only studies short-term changes extending over a few weeks.
They also attempt to convey the substantive significance of their results: "...the lethality of militant attacks declined by an average of nearly 25 percentage points in a given week in which a drone strike occurred. On average, 2.77 people were killed or injured in militant attacks in FATA between 2007 and the end of the third quarter of 2011. This figure would decline substantially to 1.73 per week as a result of a single drone strike if the number of drone strikes would increase by one per agency-week." Drone strikes were also found to be associated with a decline in militant attacks on tribal elders, which contradicts narratives about the strikes' supposedly deleterious effect on "tribal structures."
Aside: Dylan Matthews of Vox recently weighed in on this paper. He put the above numbers in perspective by arguing that, since a single drone strike is estimated to save about one person per week, even one unjustified casualty caused by the strikes, on average, will wash out the life-saving effect. Whether U.S. strikes cause more than one such casualty on average seems to depend, as with so much else, on which counts you use. But I think he has a good case that they do if you look at the entire length of the program. "Of course," he says, "it's possible that the realistic alternative to drone strikes is other military actions — like Pakistan military raids — that bring a higher death tolls, but compared to a baseline where neither the US nor Pakistan takes action, it's hard to see how the strikes save lives." I have three responses. First, even if the local life-saving effect is undone by the strikes themselves, it seems important that the U.S. may be able to accomplish its counterterrorism goals without causing more civilian casualties than would have occurred by terrorists' hands in the absence of the drone campaign. These goals (which should be questioned) likely include denying militants a safe haven for unchecked planning and training as well as limiting the militants' ability to expand further into Pakistan as they have in the past. Second, estimates from activist groups suggest that the U.S. is getting much better at avoiding civilian harm, so the life-saving balance may be improving with time. One overview of trends notes that "In 2013 the minimum number of civilians killed as assessed by TBIJ reached 0 (the maximum is assessed at 4)." Finally, when someone picks "a baseline where neither the US nor Pakistan takes action" as a morally attractive alternative, I think they need to say a few words about why that is appropriate given what is happening in the country.
Lyall on airstrikes in Afghanistan
A less optimistic but more counterintuitive story about airstrikes, both manned and unmanned, is found in Jason Lyall's paper "Bombing to Lose: Airpower and the Dynamics of Coercion in Counterinsurgency Wars." Lyall uses data on "nearly 23,000 airstrikes and non-lethal shows of force in Afghanistan (2006-11) to examine how insurgents respond to actual and threatened coercion." He finds that both airstrikes and non-lethal shows of force (!) are associated with increased insurgent attacks soon after and in areas very close to the aerial actions. This indicates that "the revenge motive" triggered by civilian casualties is likely not the pathway through which counterinsurgent violence triggers insurgent attacks. If it were, mere shows of force (like bombing runs where no bombs are released) would not produce such similar effects. In his words, "That shows of force elicit nearly the same reaction as airstrikes underscores the relative importance of reputational pressures rather than revenge motives when explaining insurgent behavior."
Lyall uses disaggregated data to advance a new explanation based on those "reputational pressures." Airstrikes and aerial shows of force give insurgents the opportunity to "build and maintain reputations for resolve" by responding with their own violence. In my understanding, insurgents can show through violence that "we're still able and willing to fight" after an airstrike, or "we're not intimidated" after a show of force. So, even though airstrikes may be useful for attrition of insurgent groups and shows of force may keep heads down, the incentives both create in favor of violence may dominate.
In contrast to Johnston and Sarbahi, Lyall finds that airstrikes are "strongly associated with net increases in the mean number of post-event insurgent attacks in targeted villages relative to control villages." Drone attacks (a subset of all aerial attacks) cause an estimated increase in insurgent attacks that is "substantially smaller than non-drone airstrikes," but this only suggests that "drones are at least less counterproductive than their manned counterparts."
How can we reconcile the results?
The findings of the two papers point in different directions: the first shows a decrease in militant attacks following drone strikes, and the second shows an increase following all airstrikes, including drone strikes.
One of the authors told me that they have discussed between them possible explanations including differences in measurement or perhaps a different conflict environment between the border areas and Afghanistan proper. I suspected at first that a measurement difference might be to blame: if one research design only counted subsequent attacks within the week after an airstrike, while the other used a longer time window, there might be inconsistent results. But the second paper tests a number of lengths and one of these matches the time window employed in the first paper, so I doubt something this simple this is the culprit. Regarding differences in conflict environment, perhaps there is some reason that the need to signal resolve and capability is more pressing for insurgents in Afghanistan than it is for militants in northwest Pakistan. By Lyall's argument, this would explain the divergence in results. But I am not sure what that reason might be.