Wednesday, October 18, 2006

Civilian casualties in Iraq: 650,000 are not dead

About a week ago, a study was published that estimated somewhere between 426,000 and 800,000 Iraqi civilians have been killed since the U.S. invasion. The mean of that number is 650,000, and for whatever reason, that number has stuck. President Bush, when asked about this study, said it had been discredited and that he didn't put any stock in the numbers. In response, various media types have gone into a frenzy, including media watchdog Eric Alterman:

According to the Johns Hopkins Bloomberg School of Public Health, George Bush's lies have killed not 30,000 innocent Iraqis, as the president not long ago estimated, but nearly 22 times that amount, or 655,000. Neither the Pentagon, nor much of the mainstream media have made much attempt to make their own counts -- it's just not that important to anyone. So how has the U.S. media reported on these shocking-albeit-necessarily-imprecise findings, based on door-to-door surveys in 18 provinces, by the experts trained in this kind of thing? The actual methods included obtaining data by eight Iraqi physicians during a survey of 1,849 Iraqi families -- 12,801 people -- in 47 neighborhoods of 18 regions across the country. The researchers based the selection of geographical areas on population size, not on the level of violence. How strict were their standards? They asked for death certificates to prove claims -- and got them in 92 percent of the cases. Even so, the authors say that the number could be anywhere from 426,000 to 800,000.

Newsflash: Bush is right, Alterman is wrong. The 650,000 figure is wrong. The 426,000 figure is almost certainly wrong. I could explain why myself, but instead I'll just steal from my friend Megan McArdle's blog (you'll need to read her entire blog to get an overall sense of her argument):

In August, according to UNAMI, there were 73 violent deaths per day in Baghdad. In July, there were 93. Extrapolating this evenly across the remaining 75% of the population would give us approximately 125,000 deaths per year in the entire country, which is still below the lower bound of the Lancet study. But we also know that the other regions of Iraq aren't equally violent. The 10% of the population that is Kurdish lives in a basically stable area. So it's really 65% of the country, or another 16.5 million people that would have a similar violence rate. That gives me roughly 110,000 total. However, we also know that the country is, on average, safer than Baghdad, though some places, like Fallujah may be more dangerous. So that number needs to be revised downwards substantially. How substantially? No idea. Finally, we also know that the violence against civilians has been getting worse; even if the collection of death certificates is shoddy, the direction of the trend is clear. So earlier years would have fewer than 116,000 deaths even if the rest of the country were just as violent (on average) as Baghdad. Call it an upper bound of 300,000, being very charitable. That's a high number. But that's the top of the range, not the most likely number. What is the most likely number? No idea, and I'm not going to join the Lancet doctors in giving a false impression of precision. But it's less than a third of the Lancet study's upper bound.

Make no mistake, far too many have died, largely as a result of poor post-war planning. But there's no way 650,000 Iraqi civilians have been killed.

***UPDATE***

Just in case the above quote didn't convince you, here's Steve Moore, a political consultant who's actually conducted suveys in Iraq.

In their 2006 report, "Mortality after the 2003 invasion of Iraq: a cross-sectional sample survey," the Johns Hopkins team says it used 47 cluster points for their sample of 1,849 interviews. This is astonishing: I wouldn't survey a junior high school, no less an entire country, using only 47 cluster points.

3 Comments:

Anonymous Nic said...

I’m not sure I trust Megan McArdle's assessment based upon her apparent knowledge of parts of Iraq that may or may not be more dangerous than others. And she uses numbers provided by UNAMI. What is to say they are more accurate or have a better methodology than the Lancet study?

Everyone who is undertaking a study has an agenda and a reason to bend the numbers.

Lies, damn lies and statistics.

10:13 AM  
Blogger Ben said...

She isn't claiming actual knowledge. Just about anyone would tell you that Baghdad is one of the most dangerous places in the country. Kurdish area, in contrast, is relatively safe. She runs her estimates without adjusting downward for "safer" areas. As far as why the Lancet study is imperfect, I'd read her entire blog posting on the subject. The Lancet study was shredded by Steve Moore in the updated quote -- you should read his whole piece too.

Your comment regarding study is oddly pointless. Of course there is an "agenda" -- to complete the study As for bending the numbers, I think you are wrong. While I wouldn't trust the tobacco companies who fund health studies, most academics make their living by accurately gathering data and forming hypothesis. Peer review and the scientific method acts as a powerful deterrent to those who would bend away.

In any event, I have no idea if Lancet has an agenda; the problem is that Iraq is so unstable, we cannot trust the data to be reliable. That's why no one should fixate on the 650,000 figure.

1:43 PM  
Blogger Johnbai3030 said...

Ben, I disagree violently. In a case like this, peer review is a joke. Academia is a joke. When money is involved (and not small amounts of it) "science" becomes something that deserves to be put in quotation marks. There aren't any scientifically valid ways to determine the total numbers of innocent Iraqi civilians who we've killed. (We can't even get accurate election results in our own countries.) The fact that we may have butchered only 100 thousand vs. 600 thousand innocent lives is the factoid that's "oddly pointless." Who fucking cares, except apologists and excuse makers. Please don't be one of those.

3:47 PM  

Post a Comment

<< Home