Number Games

Recently Lancet, the British medical journal, published a report from the Bloomberg School of Public Health at Johns Hopkins University claiming 665,000 civilian deaths in Iraq since the U. S. bombardment began in March 2003. There has been remarkably little controversy about this figure despite its being wildly out of proportion with most other data out there.
I submit the following sampling:

Iraq Body Count site counter as of last weekend: Min-- 43937 Max--48783
At least 400 civilian deaths were reported between Monday 2 and Sunday 8 October.

From Raed Jarrar's Civilian Casualties Site
The First Fifty Days Table: Casualties betw. 3-20 and 5-8/03
total injured:4959
total dead:1995
War Deaths 20 March 03-31 July 03 Table: 2066 total, boys & men 1573, girls & women 493

From BBC, October '04:
Civilian toll estimates at 10/04
Iraq Body Count: 14-16,000
Brookings Inst: 10-27,000
UK foreign secretary: >10,000
People's Kifah >37,000
Lancet: >100,000

Let's get a few things straight ere we go further. Every death, civilian or combatant, is a horrible tragedy for somebody. In the case of civilian deaths, tragedy is compounded by the fact that these were not combatants. In the best of all possible worlds, innocent bystanders aren't supposed to die. Of course, in the best of all possible worlds there would be no dictators, no extremists, and no need for military engagement in the first place. It would be nice to be able to resolve an international dispute without innocent bystanders getting hurt, but as long as there are bad guys in the world that's unlikely. Innocent bystanders make such handy bargaining chips.

Several of the studies do little to discourage the implications that all these deaths are genuinely civilian and can be laid squarely at the feet of the U. S. Armed Forces. Both assumptions are unfair, especially when it is consistently the "insurgents" who have no qualms about targeting the vulnerable and seem to seek opportunities to do so: driving explosives-laden vehicles into groups of children, raiding school buses for teenagers of the "wrong" sect, strapping explosives to the mentally handicapped, and dissembling about the combat status of their own injured.

Let's look at the Lancet numbers critically, though:
I took the total number given by Lancet, 665,000, and divided it by the number of days in a year (didn't factor in the Leap Day; sorry,) and the number of years since the beginning of hostilities. That came to 3.5 years as of mid-September. The fatality rate I got for 665,000 divided by 3.5 years comes to 190,000 deaths per year. The per day fatality rate for that figure worked out to about 520.5 (let's go easy on Lancet and just call it 520) deaths. That's 520 civilian deaths per day, seven days a week, 365 days a year, for the last three and a half years. It's also 500,000+ more than Lancet cited in 10/04, a year and a half into the war, which seems to me disproportionate enough that it ought to raise some eyebrows.

Jarrar's figures above cite 1995 deaths in the first fifty days. Most of those were during the first few weeks so an average doesn't really describe the actual situation, but 1995 divided by fifty comes to an average of 39.9 deaths per day. (If you look at his table, you will notice that the high-casualty days were up in the low hundreds, while late April and May had only a few deaths per day.) Even if we imagine that all 4959 injuries resulted in death (which would be unlikely,) that would still only come to (about) 139 deaths per day for that period. You would have to nearly quadruple the figure to get the Lancet rate for that period.
Iraq Body Count's high-end figure from its counter is 48,753 fatalities. Lancet's figure is more than thirteen times that number. People's Kifah's October '04 figure is 37,000, or about 68 per day for the first year and a half--about one-third Lancet's rate for the same period and almost one-eighth the current study's rate. The Iraq body count statistics for the past week are noteworthy. Recently, much has been made of the spike in violence. If Iraq Body count's number of 400+ civilian deaths in one week constitutes part of a spike, how can Lancet defend a figure of 500+ a day?

The Iraq the Model brothers appear fit to be tied over the Lancet study. They maintain that the numbers are not borne out by their own observations, and condemn the intent of the study:
Among the things I cannot accept is exploiting the suffering of people to make gains that are not the least related to easing the suffering of those people. I'm talking here about those researchers who used the transparency and open doors of the new Iraq to come and count the drops of blood we shed.
In the last 1000 days, Lancet is claiming an average of about 500; over 3000 every week. The reported news notes a few dozens on bad days, hundreds in the week. Where are the thousands of bodies? They're not in Iraqi hospitals or morgues -- I wonder how many death certificates have been made (and who makes them).

Over the weekend I caught part of an NPR interview with the Lancet study's director. He defended his methodology, in part, on the grounds that it had been used in other conflicts; the Sudan, for example. It's really too bad he brought that up, because it's forcing me to question the validity of the fatality rates there, as well--not something I wanted to do. If the Lancet study's researchers are allowing personal agendas to bleed over into their work, they need to consider whether the shock value of their numbers in this one study is worth the potential damage to any good work they have previously done.