The U.S. withdrawal from Afghanistan this past August brought an end to a 20-year war. But as a series of recent investigations by The New York Times has underscored, it also marked the beginning of postmortems about what the United States did right and, in some cases, did wrong.
Drawing on Pentagon documents obtained through the Freedom of Information Act, the Times revealed that U.S. drone strikes killed an alarming number of civilians in Afghanistan—likely hundreds more than the 188 the Defense Department has acknowledged killing in such strikes since 2018—a pattern that appears to be consistent with U.S. operations in Iraq and Syria. Targeting decisions were sometimes marred by confirmation bias: Pentagon analysts saw what they expected to see, often identifying civilians rushing to help those hit by U.S. strikes as terrorists and striking them, as well. This reporting is an important first step toward accounting for the shortcomings of the drone war that Washington launched after 9/11, one that President Joe Biden’s administration should build on as it concludes its own review of drone strikes outside conventional—or declared—war zones.
But no accounting of the drone war would be complete without determining whether policies intended to reduce civilian casualties from U.S. strikes ever worked. To answer that question, we studied strike data from Pakistan, where the Pentagon and the CIA reportedly conducted nearly 400 strikes in the ten-year period before President Barack Obama’s administration tightened its targeting requirements. In 2013, the administration officially shifted its standard from “reasonable certainty” of zero civilian casualties to “near certainty.” Our analysis shows that this policy change dramatically reduced civilian casualties in Pakistan without giving terrorists an appreciable advantage, suggesting that similarly stringent targeting standards might save innocent lives in theaters such as Iraq and Syria, too.
IN PURSUIT OF PRECISION
According to the nonprofit Bureau of Investigative Journalism, which collects data from news reports, official statements, press releases, and other documents, U.S. drone strikes in Afghanistan, Pakistan, Somalia, and Yemen from 2002 to 2020 killed between 10,000 and 17,000 people. Between 800 and 1,750 of the dead are thought to have been civilians, the highest percentage of them in Pakistan. But these aggregated data mask differences in the rate of killing over time—and in the criteria that different U.S. administrations used to balance the need to target suspected terrorists with the requirement to protect civilians.
The last year of President George W. Bush’s administration saw a tenfold increase in strikes from the previous three years combined. The Obama administration accelerated this trend, conducting three times as many strikes in its first two years as the Bush administration conducted in its entire second term. The accompanying surge in civilian casualties, which amounted to three civilian deaths per strike in 2009, drew criticism from the United Nations and from watchdog groups such as Amnesty International. It was against this backdrop that Obama adopted a set of more stringent requirements for U.S. strikes in undeclared theaters, including Pakistan.
Critics have suggested that the policy achieved little. The legal expert Jameel Jaffer argued in his 2016 book, The Drone Memos, that Obama’s “stricter limitations on drone strikes seems to have had an effect only at the margins.” Such analysis is based on overall civilian casualty figures, however, rather than the rate of change in civilian deaths over time. It also does not look at the mechanisms through which Obama’s near certainty standard may have affected results on the ground. By 2017, moreover, the question of the effectiveness of Obama’s policy appeared moot, since President Donald Trump returned to the more permissive reasonable certainty standard.
U.S. drone strikes have killed an alarming number of civilians in Afghanistan.
Our analysis takes a different approach, focusing on the effect Obama’s near certainty standard had on the rate of reported civilian casualties following U.S. drone strikes as well as the rate at which strikes hit their intended targets. Beginning in 2011, U.S. officials began debating more restrictive targeting guidelines, and Obama appears to have begun ratcheting up requirements for strike approvals in undeclared conflict zones—namely, Pakistan. Unlike U.S. strikes in Afghanistan and Iraq, which continued under the more permissive reasonable certainty standard because they were often in defense of U.S. soldiers, strikes in Pakistan began to be conditioned on near certainty of no collateral damage. Obama’s ultimate goal, senior officials responsible for drafting, implementing, and auditing the reform told us, was to encourage commanders and intelligence officials to conduct more precise strikes, which would also help rehabilitate the United States’ image abroad following the public outcry over civilian deaths.
Almost immediately after the administration began its policy deliberations in 2011, civilian casualties from U.S. strikes in Pakistan markedly decreased. On average, our analysis of strike data from the Bureau of Investigative Journalism reveals that the deliberations and subsequent policy shift resulted in a reduction of 12 civilian casualties per month or two civilian casualties per strike. In doing so, the policy increased the precision of U.S. strikes—the proportion of casualties that were the intended targets of strikes—to 95 percent in Pakistan. In other words, under Obama’s policy, U.S. strikes in Pakistan approached near perfect accuracy.
Had Obama not changed the policy, we estimate that U.S. strikes in Pakistan would have resulted in several hundred more civilian deaths between 2011 and the end of his second term in 2017. The decline in civilian casualties is more than a statistical artifact: it followed a deliberate effort by the Obama administration to impose stricter targeting procedures, including delegating fewer strike decisions to commanders in the field who, according to The New York Times, often misidentified combatants due to cognitive bias. Obama’s policy was explicitly designed to inculcate a culture of “effective and discriminate use of force” to overcome such bias, which can result from battlefield pressures such as anxiety, fear, and the desire to protect friendly forces even if doing so means putting civilians at greater risk.
The case of Pakistan shows that a standard of near certainty for U.S. drone strikes can reduce civilian casualties. But such a stringent standard imposes tradeoffs. Moving targeting decisions up the chain of command can result in longer approval timelines for strikes. One policy analyst we interviewed cautioned that Obama’s policy resulted in “missed opportunities” to take out terrorists. But our research shows that instances in which time is truly of the essence are rare. The typical strike benefits from “red teaming,” a practice that involves bringing in an outside group of analysts and commanders to question the merits of a strike based on the full body of intelligence, taking into account any information gaps. The fact that there were no major terrorist attacks on U.S. soil during Obama’s presidency suggests that his higher targeting threshold in Pakistan did not come at an appreciable cost to national security.
A tighter threshold for U.S drone strikes can reduce civilian casualties without emboldening the enemy.
If a higher targeting threshold succeeded in reducing civilian casualties in Pakistan without endangering Americans, there is reason to believe that similar targeting adjustments could yield the same results in other countries where the United States is currently conducting drone strikes. One complicating factor is the presence of significant numbers of U.S. forces on the ground. In declared theaters of operation, strikes are often conducted in support or defense of U.S. soldiers. As a result, critics of the tighter targeting standards caution that these standards could expose U.S. soldiers to greater battlefield risks, since commanders could be restricted in their ability to use drone strikes for close air support. This tradeoff is less pronounced in undeclared theaters of operations, where there are fewer U.S. boots on the ground. Still, the Biden administration will have to strike the right balance between military necessity and protecting civilian lives in its ongoing review of the United States’ drone program.
Regardless of the specific theater of operations, limiting civilian casualties in war should be an end in itself, not just because the United States is bound to do so under international law but because civilian casualties are thought to make it easier for terrorists to recruit followers. Our analysis suggests that reducing civilian deaths from U.S. drone strikes doesn’t have to come at the cost of effective counterterrorism. A tighter threshold for U.S drone strikes can reduce civilian casualties without emboldening the enemy.