Targeted Killing: Drones

Civilian Deaths from Drone Strikes

By Benjamin Wittes
Friday, August 12, 2011, 3:08 AM

Scott Shane has an important story in the New York Times on the controversy over the number of civilian deaths caused by the CIA's drones program outside of Afghanistan. Shane reports that,

The civilian toll of the C.I.A.’s drone campaign, which is widely credited with disrupting Al Qaeda and its allies in Pakistan’s tribal area, has been in bitter dispute since the strikes were accelerated in 2008. Accounts of strike after strike from official and unofficial sources are so at odds that they often seem to describe different events.

The debate has intensified since President Obama’s top counterterrorism adviser, John O. Brennan, clearly referring to the classified drone program, said in June that for almost a year, “there hasn’t been a single collateral death because of the exceptional proficiency, precision of the capabilities we’ve been able to develop.” Other officials say that extraordinary claim still holds: since May 2010, C.I.A. officers believe, the drones have killed more than 600 militants — including at least 20 in a strike reported Wednesday — and not a single noncombatant.

Cutting through the fog of the drone war is important in part because the drone aircraft deployed in Pakistan are the leading edge of a revolution in robotic warfare that has already expanded to Yemen and Somalia, and that military experts expect to sweep the world.

. . .

The government’s assertion of zero collateral deaths meets with deep skepticism from many independent experts. And a new report from the British Bureau of Investigative Journalism, which conducted interviews in Pakistan’s tribal area, concluded that at least 45 civilians were killed in 10 strikes during the last year.

Others who question the C.I.A. claim include strong supporters of the drone program like Bill Roggio, editor of The Long War Journal, who closely tracks the strikes.

“The Taliban don’t go to a military base to build bombs or do training,” Mr. Roggio said. “There are families and neighbors around. I believe the people conducting the strikes work hard to reduce civilian casualties. They could be 20 percent. They could be 5 percent. But I think the C.I.A.’s claim of zero civilian casualties in a year is absurd.”

As Shane later reports, Brennan has modestly backed off his claim, stating now that the government does not have "credible evidence" of a single civilian death. But that distinction will not quell the controversy, about which a few thoughts are in order.

The range of dispute here is remarkably wide. If the number of civilian deaths is really as low as the CIA claims--roughly 50 out of more than 2,000 kills since 2001--it represents a less-than 2.5 percent rate of collateral killing. This would be a historic accomplishment in warfare and an enormous cause for celebration on grounds both of combat effectiveness and humanitarian concern for civilian protection. By contrast, according to the report Shane cites from the Bureau of Investigative Journalism, the number of civilians killed is somewhere between 385 and 775. This is out of a total of "Between 2,292 and 2,863 people . . . reported to have died in the attacks – most of them militants." That implies a rate of collateral deaths somewhere between 13 percent and 34 percent--itself a wide disparity. Now, to be sure, a one-third rate of collateral death is on its own terms quite low by the horrific historic standards of air power. But it is a far cry from the almost-pristine CIA estimates--which, over the past year, are quite literally pristine--or even from the impressive lower-range collateral estimates given by the Bureau report. Speaking personally, I would sleep far more comfortably knowing that the agency has a 13 percent rate of accidental death--which seems to me pretty good--than I would knowing that one third of its kills have involved people who were not, in fact, the enemy. In other words, the debate on which Shane is reporting here really matters in terms of the way we will--and should--feel about this program.

Unfortunately, I suspect the dispute is--at least in the short-term--irresolvable. Waziristan is, as Shane reports, a black hole information-wise. There will be entirely legitimate disputes--given the enemy's propensity to mingle with the civilian population--as to whether individuals killed are, in fact, civilians or fighters. The CIA will not release the kind of information that gives people like Brennan the confidence to make statements like the ones he has given. And whether because of legitimate skepticism of government action or because of active hostility to U.S. counterterrorism operations (both factors are at work, depending on the group), critics and the press will often err on the side of giving credence to claims of civilian casualties. It is hard to imagine a credible public voice with access to enough good data to resolve this dispute in a fashion that will win general acceptance.

This means that we are, at some level, inevitably proceeding on the basis of an instinct about how much we trust government to know and report its error rate. And ironically, this might be easier were the rate itself a little less perfect. There's something hard to believe about a magic program that removes terrorists--and nobody else--from the face of the earth. I'm not suggesting that the CIA make up some civilian casualties to bolster its credibility. But I do wonder if the manner in which Brennan is reporting the numbers magnifies the sense of certainty that innocents haven't been killed over and above what we really know. In Shane's story, Brennan says that “Fortunately, for more than a year, due to our discretion and precision, the U.S. government has not found credible evidence of collateral deaths resulting from U.S. counterterrorism operations outside of Afghanistan or Iraq." I wonder if this number would look different if Brennan reported the number of deaths the U.S. could not confirm were militants--in other words, reported the number of deaths that, in its judgment, might be civilian.