Thursday, October 26, 2006
A possibly final-ish note regarding the Lancet report
One of the complaints about the Lancet report was that it didn't match up with two other estimates: the Iraq Body Count estimate, and the Iraq Living Conditions Survey.
The Iraq Body Count estimate is known to be low, so, while it shouldn't be ignored, the real challenge is the ILCS, which was conducted in a manner similar to the two studies published in Lancet.
The ILCS suggested there were 24,000 war related deaths in 2004. The Lancet article in 2004 suggested the number was 100,000. How could this be, especially when the ILCS had a much higher number of clusters?
Simple. The ILCS covered 12 months post invasion. The Lancet study covered 18.
"Okay," you might say, "but that's not enough! That would make the Lancet study possibly correct if it suggested there were 36,000 deaths due to the war, not 100,000!"
Ah... but there are two relevant points to consider.
First, the most recent survey clearly shows that the death rates increased each year, so violent deaths were occurring more frequently as time went on.
"Okay, so, let's just be really generous," you might say, "And let's pretend that the number of war related deaths was 50% greater in the next six months. That'd be 42,000 deaths, not 100,000!"
Which brings us to the second point. The studies in Lancet all counted excess deaths, due to all causes. When you look at it's number for war related deaths, you come up with...
39,000. (With thanks to Tim Lambert.)
Even if you assumed the death rates stayed exactly the same as in the ILCS, you'd expect 36,000, and the 39,000 estimate had a big enough margin of error that it included 36,000 as a pretty good possibility as well.
In short, the ICLS seems to support the Lancet study. (Or, to put it another way, the Lancet study seemed to support the ICLS.)
The ICLS might be more accurate, but its numbers do not call the Lancet numbers into question.
He has some interesting things to say about the 2004 paper and how it compares to other estimates. I seem to recall one of them involved interviews of military personnel...but I can't be sure. I doesn't seem to be downloading at the moment. Perhaps it will shortly.