Check some notable reactions to the latest WHO/IFHS estimate of Iraq mortality, due to violence, from Les Roberts, Eli Stephens, Juan Cole, Kevin Drum, Daniel Davies, Andrew Cockburn, Naomi Spencer, and Lenin's Tomb.
This comment from Lenin's Tomb is particularly worth noting, being that it suggests the potential for a kind of self-reinforcing underestimate of deaths. A substantial one, frankly:
One point of serious divergence with the 2006 Lancet survey is that this study attributes over 50% of the deaths to Baghdad and only around 30% to "High-Mortality Provinces". The Lancet survey attributed about 20% of the violent deaths to Baghdad and over 60% to those "High-Mortality Provinces". One reason for this appears to be that of the clusters they could not visit due to "security" problems (bombing and shit), 61.7% were in Anbar. Thus, to make up for it, they calibrated the figures according to the Iraq Body Count database - the database that, as we all know, is compiled from media reports, which of course are themselves blind to huge parts of the country due in part to "security" risks. Further, it is not even based on all media reports of deaths: about one in four stories in the major US press alone have been missed by the IBC's survey according to one review (cited by Roberts and Burnham). Thus, their distribution of 'reported' deaths is very much like IBC's. This distribution is counterintuitive since the largest US military operations have been conducted in towns and cities in the 'Sunni Triangle' such as Fallujah, Ramadi, Haditha and al-Qaim - all in the Anbar province. The biggest operations, such as Operation Phantom Fury, were focused in Anbar. The fiercest resistance to the US has been based in Anbar, with the province accounting for the largest portion of attacks of all kinds in each of the three years covered.More or less, this is an admission that the most violent areas of Iraq were effectively excluded from the study. And to the extent that the IFHS team tried to reconstruct the mortality rate from afar using passive figures (IBC's), they're compounding the undercounting bias since passive figures for particularly violent and anarchic regions are especially subject to pressures that tend to lead to underestimation -- most notably, a limited media presence and severely degraded public health monitors (morgues, hospitals, regulated burials, death certification from the state, etc.).
That said, the WHO/IFHS study is not worthless. Far from it. It just needs to be contextualized properly, like all the rest of the mortality estimates. Unfortunately, that rarely happens.