Bush disputes its validity, claiming that its methodology is flawed. The Iraqi government denies it, claiming fewer Iraqi deaths have occurred than even indicated by surveys based primarily on obviously incomplete coroner's records.
Both, however, are intentionally deceiving their people.
The methodology in this case appears to be excellent. All the surveying was conducted by teams of medical doctors, capable of speaking both Arabic and English, using the same methodology as used and accepted in Chad and Darfur. The extent of the survey was enormous, with 12,801 individuals in 1,849 households personally surveyed. The degree of evidence requested by those surveyed was also excellent, with families of deceased Iraqis providing death certificates in over 90 percent of cases.
The U.S. regularly accepts poling data of one person for every 285,000 as being valid, plus or minus 5%. In this case, the polling data was approximately one person for every 2,170 Iraqis. That's extremely significant. A far larger population was surveyed than in the prior Lancet report, and the results obtained were extremely consistant with the prior report.
Also note what the survey measures -- excess morbidity above statistically expected levels prior to the war. That means Iraqis who die from terrorism, from internal violence, from fighting coalition troops, from disease, malnutrition, cancer, lack of access to emergency medical aid, etc.
Those numbers could easily number in the hundreds of thousands, as I've argued in the past.
And no, approximately 655,000 dead Iraqis is absolutely not an indictment of the inhumanity of our troops. Instead, it is ultimately an indictment of the inhumanity of war, and of those who brought our nation and the people of Iraq in to war, without providing us a clear way out.
As a way of comparing the increased level of accuracy of the updated Lancet report vs. the original one, it's important to pay attention to the statistical confidence interval
The old Lancet report was highly criticized for having a smaller number of surveyed households, which led to a 95% confidence interval that, although very likely to be close to accurate, was only 8000 deaths at one extreme, and 194,000 deaths at the other, for the first 13 months of conflict. The new survey, however, provides a far more significant, accurate statistical range between 430,000 and 710,000.