The Iraqi Death Toll

This latest estimate of 655,000 Iraqi deaths due to the war seems preposterous, but the methodology seems sound. For a good summary of how the figure does not pass the smell test, see the reader comments in Steve Sailer's commentary. Steve himself seems inclined to accept the figure because the methodology is sound. But the reader makes a helluva point that the study essentially claims 80,000 car-bombing victims, or 135 a day . We know that's out of bounds, since car-bombings are usually reported in the news, and make real big news when the death toll exceeds a couple dozen. Randall Parker also discusses the survey, and in the comments Laurence Auster points out that this number of dead is about the same level as U.S. Civil War dead - a war which featured major pitched battles where 10,000 would die in a single day. Nothing like that has ever occurred in this war (though they might have been common in the Iran-Iraq war).

This is how the study was conducted:

A sample size of 12000 was calculated to be adequate...to balance the need for robust data with the level of risk acceptable to field teams...[S]election of survey sites was by random numbers applied to streets or blocks rather than with global positioning units (GPS), since surveyors felt that being seen with a GPS unit could put their lives at risk...By confining the survey to a cluster of houses close to one another it was felt the benign purpose of the survey would spread quickly by word of mouth among households, thus lessening risk to interviewers.

As a first stage of sampling, 50 clusters were selected systematically by Governorate [province] with a population proportional to size approach, on the basis of the 2004 UNDP/Iraqi Ministry of Planning population estimates. At the second stage of sampling, the Governorate's constituent administrative units were listed by population or estimated population, and location(s) were selected randomly proportionate to population size. The third stage consisted of random selection of a main street within the administrative unit from a list of all main streets. A residential street was then randomly selected from a list of residential streets crossing the main street. On the residential street, houses were numbered and a start household was randomly selected. From this start household, the team proceeded to the adjacent residence until 40 households were surveyed.

So one obvious problem is that the sampling is not really random since, ironically, the situation is just too dangerous to do the job right. Thus while they sampled some 1800 households, they were concentrated in 47 clusters (i.e., neighborhoods - once they randomly chose the neighborhoods, they then went door to door within each neighborhood to find 40 households). That the survey could have oversampled in more dangerous neighborhoods doesn't seem too farfetched to me.

The raw data is frustratingly not available on the Lancet site. It would be very interesting to look at the distribution of deaths by neighborhood and household within neighborhood. The calculations are also not detailed, but the basic approach doesn't seem too much more sophisticated than extrapolating the survey deaths proportionately out to the Iraqi population as a whole. It would be better I think to take into account relative population densities of the selected clusters vis a vis the surrounding geographic region than simply adding up all the survey deaths (as if the sampling were truly random) and extrapolating the total.