In an interview today with Alex Sarotin of Radio Free Europe I discussed the dispute over The Lancet’s publication of a Johns Hopkins Bloomberg School of Public Health Study that the number of fatalities in Iraq attributable to the war is substantially higher than the 30,000 cited by the Bush administration. The 655,000 number bandied about in the press is a kind of upper limit that is probably excessive because it includes data from Fallujah, which is probably not representative of Iraqi society as a whole because of the unusual intensity of the violence there. However, excluding Fallujah, the research team claims there are approximately 98,000 additional deaths attributable to the war, still more than three times that quoted by the Bush administration. What’s the story?
The research here was a random sampling of household clusters which was mathematically extended to the country as a whole. There are three main sources of error with which we must be concerned: (1) systematic error due to the fact that the sampling is not truly representative; (2) random errors due to the fact that the study is statistical rather than a comprehensive survey and can never be as accurate as a comprehensive survey; and (3) blunders by the research team, i.e., the possibility that the research team simply did not do its job right.
The authors of the Lancet article discuss the systematic errors in great detail. As mentioned above they properly caution about the Fallujah cluster, although the press naturally has ignored that when writing their shocking “655,000 deaths†headlines. However, there are other systematic errors that imply the 98,000 figure is an underestimate. For example, the fact that they sampled by interviewing households means that those households that were completely destroyed are not included. Being totally destroyed they could be a significant factor. Further the fact that the 98,000 figure excludes Fallujah means that it will be an underestimate.
The authors also provide the range of random error due to the fact that their study is statistical. While estimating that the death rate (excluding Fallujah) after the invasion as one and a half times what it was before the invasion, the authors not that here is a 95% chance that the actual value is somewhere between 1.1 time and 2.3 times higher. This means that there is a 2.5% chance that in fact the Bush administration estimates are not far off; but there is 2.5% also that the excess deaths are more than 194,000.
The last concern is the ability of the researchers, but that is not seriously in question here. The team is affiliated with a top-notch institution and the paper was peer reviewed four scientists before acceptance by The Lancet. The bottom line is that there have been many civilian deaths due to the invasion of Iraq. Every one recognizes this.
Imad-ad-Dean Ahmad
Leave a Reply