Army-Funded Study: We Have No Idea If Drones Work


A meta-analysis of the research on drone strikes in Afghanistan and Pakistan published on Thursday, funded by the U.S. Army War College, found very little evidence that the targeted killing campaign was successfully degrading al-Qaeda and the Taliban. It even found some reason to believe that that the strikes were backfiring — but not for the reasons you might think.

James Igoe Walsh, a professor of political science at the University of North Carolina-Charlotte, wrote the paper for the Strategic Studies Institute, the policy research branch of the College. Walsh’s paper surveys existing regression analyses — the most mathematically rigorous work on the drone campaign — to see if they came to any consensus about the effectiveness of the targeted killing program.

Because the data on the drone campaign are relatively new, Walsh only found four such papers worth surveying (two of which he coauthored). Here’s what they said.

First, “one reasonably consistent finding across the spectrum of analysis is that drone strikes have little influence, positive or negative, on the amount of insurgent violence that occurs in Afghanistan.” Not one of the four papers found much of a correlation between drone strikes and violence and Afghanistan, suggesting that if a goal of the campaign is to reduce Al Qaeda and the Taliban’s ability to kill American troops and Afghan allies, it probably isn’t working.

The picture in Pakistan is more muddled. “In most of the models reported here,” Walsh writes, “drone strikes are associated with both an increase and a decrease in subsequent terrorism.” Indeed, two methologically sound papers — one by Patrick B. Johnston and Anoop K. Sarbahi and another by David A. Jaeger and Zahra Siddique — came to opposite conclusions on this question.

It’s possible, as Walsh suggests, that both are true: attacks in Pakistan decrease in the short run because they kill the people planning to carry them out, but increase in the long run by inciting anti-American anger that helps extremist groups recruit. But the different models come to different conclusions on the timing of attacks. We simply don’t have good enough models to figure out the effect of drone strikes on levels of violence in Pakistan.

A third “more tentative” finding sheds more light on the situation. “Drone strikes that result in civilian deaths,” Walsh reports, “appear to have little relationship with subsequent insurgent violence.” This one’s rather surprising. The most prominent argument for why the drone campaign might be counterproductive is that strikes cause “blowback;” that civilian casualties inflame the Afghan and Pakistani public, driving recruits into the arms of the very organizations the strikes are trying to defeat.

However, none of the regressions found that drone strikes that killed higher numbers of civilians correlated with higher levels with insurgent violence than observed after more discriminate drone strikes. So there’s not much evidence for the idea that “greater numbers of civilian casualties mean more support for terrorists.”

There’s another way that drone strikes might by backfiring. In Walsh’s paper with John Szmer, “They Must All Be Militants: Targeted Killings, Drone Strikes, and Insurgent Violence in Afghanistan and Pakistan,” they find no correlation between civilian death and subsequent insurgent violence, but a positive relationship between insurgents killed and increased violence. In other words, the more drone strikes hit their actual targets, the less effective they are.

Though this may sound ridiculous, it makes a certain kind of sense. What we call “the Taliban” is actually a loose agglomeration of different groups with different priorities and objectives. Though drone strikes are more precise than virtually any other U.S. aircraft, they still aren’t precise enough to tell if a specific target is affiliated with, say, the Haqqani Network or the Quetta Shura Taliban. If the same bombing campaign is targeting all of these groups, the theory goes, they’re more likely to coordinate with each other in the face of a common enemy, increasing their overall effectiveness. There’s some evidence from the Vietnam War supporting the basic theory.

Walsh’s conclusion: “drones appear to be, at most, weak substitutes for traditional counterinsurgency operations.” They’re not, on their own, going to seriously weaken the Taliban. While drone strikes might be more effective if the sole goal were to degrade al-Qaeda’s ability to attack the American homeland, Walsh finds that “this claim is difficult to assess, however, since the United States has not employed drones consistently in a counterterrorism campaign.” Instead, “the United States has tended to expand the targets of drone strikes from individuals who appear to be planning attacks on the U.S. homeland, close allies, or forces in Afghanistan.”

Between 2004 and 2011, U.S. drone strikes in Pakistan killed between 93 and 638 civilians, depending on which independent methodology for counting casualties you accept.