Incoming GWU Ph.D. student Katherine Tiedemann has an “op-ed”:http://www.nytimes.com/2010/04/26/opinion/26bergen.html?ref=opinion in the _New York Times_ with Peter Bergen.
bq. A survey we have made of reliable press accounts indicates that since January 2009, the reported strikes have killed at least 520 people, of whom around 410 were described as militants, suggesting that the civilian death rate is about 20 percent. It’s possible, however, that the number is even lower. An American counterterrorism official told The Times in December that the civilian fatality rate is only 5 percent, saying that “just over 20” civilians and more than 400 militants were killed in 2009. Should the American government’s claims about the small number of civilian deaths be verified, some of the Pakistani hostility toward the United States might dissipate. This would be much easier if the now-classified videotapes of drone strikes were made available to independent researchers.
This speaks to the problems of gathering good data on war time casualties. It is at the least plausible that the estimate by the American counterterrorism official is a serious undercount – and without information on the methodology (if any) that he or she is basing this claim on, we can’t assess it. In general, it is probably not a good idea to rely on civilian casualty numbers from any actor which has a systematic incentive to undercount (the US government) or overcount (insurgents) them. The methodology adopted by Bergen and Tiedemann seems a plausible one – but may too result in undercounts if not all civilian deaths make it into the media which the researchers consider to be reliable (or overcounts if spurious information makes it into the press). This is plausibly less a problem than in Iraq (where the similar Iraq Body Count methodology has come under some fire), but may still be an issue (information may only percolate indifferent well out of the tribal areas to newspapers). Cluster analysis is probably not a viable approach given the relatively small number of deaths as a proportion of the population, and the obvious difficulties in carrying out research in this region. All in all, it is important that we have some independent information – and Bergen and Tiedemann’s approach seems the best available under the circumstances.
