Civilian Casualty Rates in CIA vs. DOD Drone Programs: Apples and Oranges

Much attention has been given to whether CIA-directed or DOD-directed drone strikes result in fewer civilian casualties. Last week I wrote about the mixed evidence when it comes to making that assessment. Here I want to take up a different dimension of this topic: whether such comparisons even work at a basic level if the CIA and DOD programs include (1) different target lists and (2) different operational environments.

Some of the factors that I discuss in this post cut in favor of the CIA program—they suggest that the CIA’s civilian casualty rates are even better than one might otherwise think. And other factors cut in the exact opposite direction—favoring the DOD’s performance. Taken as a whole, these factors show the extraordinary hazards in trying to make simple comparisons. Understanding these factors can also help identify the more fine-grained information that would benefit public discussion and oversight of drone operations. And finally, these factors also show how some policy suggestions such as “allowing CIA and DOD to compete over drone strikes” (see also Greg McNeal’s suggestion p. 787) would have to grapple with problems of incommensurability.

Here’s the upshot: The CIA program reportedly involves a much shorter target list and focuses more narrowly on HVTs (High Value Targets). In their book, Top Secret America, Dana Priest and William Arkin write, “Intelligence officials involved in the CIA selection process say there were never more than two or three dozen individuals on the list at one time.” Compare that to JSOC’s target list. In his book, Jeremy Scahill writes that under General Stanley McChrystal, “by October 2009, there were more than 2,000 people on the Joint Prioritized Effects List” in Afghanistan. Priest and Arkin go on to provide a direct comparison of the two lists:

“The CIA’s targeted killing campaign, no matter how successful, paled in comparison to the size of the drone war being waged abroad by the U.S. military, mainly through JSOC and mostly in Afghanistan. JSOC’s list of people to kill was much longer and more fluid than the CIA’s …. This is because the military is allowed— encouraged, even— to capture or kill all the people involved in an identifiable network of terrorists, not just its leaders. The military has ‘a lower high bar,’ as one commander put it, for putting an individual on the list and for being able to kill or capture all of his associates, if they can be found.”

If these reports are generally accurate, there are two implications for analyzing civilian casualty rates across the two programs:

1. Collateral damage assessments

Comparisons of the two programs may depend on how you think about “accuracy.” The leading organizations that track civilian casualties—the Bureau of Investigative Journalism and the New America Foundation—do not differentiate among different types of civilian casualties. They just tally the total number of civilians killed. But not all of those deaths are due to inaccurate strikes. Indeed, a commander faithfully applying the law of armed conflict might conduct some of those strikes all over again in the exact same way despite the effect of civilians. Because she did nothing wrong.

Here’s why: One reason for civilian casualties is due to errors – in killing the wrong person or in assessing the incidental risks to nearby civilians.  But another is no mistake at all – rather it is collateral damage that the law of armed conflict accepts as long as it is proportional to the value of striking a legitimate military target. The higher the value of the military target, the greater risks to civilians permitted by the law of armed conflict.

Because the CIA’s target list is heavily weighted in favor of HVTs, it is understandable that the CIA program might result in a greater proportion of civilian casualties. And if the CIA track record involved a low rate of civilian casualties, it would be even more noteworthy that the agency is able to accomplish such results given its focus on HVTs.

For illustrative purposes, here’s another way of putting it. Imagine the following:

10 CIA drone strikes result in 10 HVTs killed and 10 civilians killed

10 DOD drone strikes result in 10 lower value targets killed and 9 civilians killed

In that hypothetical scenario, the CIA’s performance may be much better—at least as far as the laws of war is concerned—than the DOD’s.

2. Targeting lower-value militants

If the reports of the two target lists are correct, it may mean the DOD is carrying out more difficult drone strikes. Lower-level militants may be more difficult to identify. Their “patterns of life” may not be as readily distinguishable compared to senior-level commanders, and they may be more intermingled with civilian life. As an example, Greg McNeal’s study suggests that the overwhelming reason for civilian casualties in DOD pre-planned operations is due to mistaken identification (“70 per cent of the time it was due to failed ‘positive identification’ of a target”).

3. Operational environment

Finally, it would be a mistake to leave the topic of comparability without mentioning that there are obvious analytic problems with comparing (CIA-directed) drone strikes in Pakistan against (DOD-directed) drone strikes in other countries. Cross country variations include differences in demographics, in US relationships to local security forces, in US ability to gather post-strike information on the ground, and in the general quality and quantity of SIGINT and HUMINT. One operational environment may pose more challenges (or present more options) for military targeting and can’t easily compare to other environments. CIA-directed versus DOD-directed drone strikes in the same country (e.g., Yemen) would offer a more useful comparison. The problem is when commentators and US officials make comparisons across operational environments or don’t tell us that’s an assumption baked into their analysis.

 * * *

These factors naturally implicate how we think about whether drone operations should shift to greater DOD control. They also suggest more generally how discussions in the United States might evaluate the performance of targeting operations conducted in our name. 

About the Author(s)

Ryan Goodman

Co-Editor-in-Chief of Just Security, Anne and Joel Ehrenkranz Professor of Law at New York University School of Law, former Special Counsel to the General Counsel of the Department of Defense (2015-2016). You can follow him on Twitter @rgoodlaw.