Detailed information from the families of those killed in drone strikes in Pakistan and from local sources on strikes that have targeted mourners and rescue workers provides credible new evidence that the majority of the deaths in the drone war in Pakistan have been civilian noncombatants - not "militants," as the Obama administration has claimed.

The new evidence also shows that the statistical tally of casualties from drone attacks in Pakistan published on the web site of the New America Foundation (NAF) has been systematically understating the deaths of large numbers of civilians by using a methodology that methodically counts them as "militants."

The sharply revised picture of drone casualties conveyed by the two new primary sources is further bolstered by the recent revelation that the Obama administration adopted a new practice in 2009 of automatically considering any military-age male killed in a drone strike as a "militant" unless intelligence proves otherwise.

The detailed data from the two unrelated sources covering a total 24 drone strikes from 2008 through 2011 show that civilian casualties accounted for 74 percent of the death toll, whereas the NAF tally for the same 24 strikes showed civilian casualties accounted for only 30 percent of the total.

The data on 11 drone strikes from 2008 through 2011 were collected in 2010 and 2011 from families of victims of the strikes by Pakistani lawyer Mirza Shahzad Akbar. Those 11 cases represent only a fraction of the total number on which Akbar has obtained data from victim's relatives.

Although relatives of drone strike victims could have a personal interest in declaring the innocence of their relatives, the details provided by relatives in legal affidavits, such as the age, employment and other characteristics of the victims, appear in almost every case to support their claims that those killed were not actively involved with al-Qaeda or other military organizations.

The data on 13 drone strikes targeting rescuers and mourners from 2009 through 2011 were gathered by the London-based Bureau of Investigative Journalism (BIJ) in a three-month investigation in late 2010 and early 2011 involving interviews with eyewitnesses and others with direct knowledge of the strikes.

The NAF Database: Origins of the Discrepancies

The NAF "Year of the Drone" project, headed by terrorism expert Peter Bergen, has been tracking casualties from drone strikes in Pakistan and estimating casualties from the strikes since 2009 based on news media reports.

But Bergen's estimates are not focused on estimating civilian casualties. Instead, they track the deaths of an undefined category of victims called "militants" by individual drone strike and by year. Those totals are shown in graphs with the residual category of "other" reflecting the overall total minus the total of "militant" deaths.

A major problem with the NAF statistics on drone victims is the extraordinarily wide spread between the low and high estimates for total number of deaths from drone strikes, as well for as the total number of "militants" killed. The range in the total number killed in strikes is estimated in the NAF database at a low of 1,879 and a high of 2,887. The NAF estimates the "militant" deaths from a low of 1,586 to a high of 2,416.

Bergen deals with high-end estimates that are 54 and 52 percent above the low end estimate by averaging them out. But the real issue is whether a very large proportion of the dead referred to by those anonymous sources giving the totals to reporters in Pakistan as "militants" were, in fact, noncombatant civilians.

The data compiled by Akbar and the BIJ strongly suggest that conclusion.

The NAF tally on the 11 strikes on which Akbar collected data from the victims' families shows a total of 66 to 78 "militants" killed along with 39 to 47 "others" - the term NAF uses in place of "civilians." But the information from the victims' families indicates that the number of "militants" killed was actually 27 to 34, while the number of civilians killed was 86 (See addendum below).

Instead of representing only 30 percent of the total casualties in those 11 strikes, as portrayed in the NAF accounting, civilian casualties actually accounted for nearly three-quarters of the total, according to the relatives' testimony.

The data on 13 drone strikes targeting funerals and rescue efforts reported by the BIJ in February similarly contradict the NAF tally of deaths. The NAF recorded a total of 90 to 176 dead in 12 strikes which the BIJ was able to confirm as targeting rescuers or mourners; 77 to 153 of the dead were listed as "militants," whereas only 13 to 24 were listed as "civilians." But eyewitnesses and other sources considered reliable in the localities reported that between 80 and 107 civilians had been killed in these attacks on rescuers or mourners. That suggests that the higher estimates for "militants" usually included the civilians killed in those strikes.

So, when adjusted for the new data, the estimate of "militants" killed would be 77 to 112, and the figure for civilians would be 80 to 107. The revised total of civilian deaths in those strikes is essentially equal to the revised total for "militants."

Combining the data on the two sets of drone strikes, the original estimate for "militant" deaths in the NAF accounting was a range of 143 to 231, but the figure based on actual local testimony is 104 to 146 - a 60 percent decrease. The figures for civilian deaths, on the other hand, increases by 66 percent, from the range of 52 to 71, based on the NAF tally, to an adjusted range of 164 to 193.

Thus civilian casualties, which were less than a third of the "militant" casualties in the NAF accounting for the 24 drone strikes in question, are revealed to be 70 percent of the total.