Friday, October 29, 2004

100,000?

The Guardian is reporting that The Lancet has published a study on its wedsite estimating Iraq deaths as a result of the war to be 100,000.

I can't find the article on The Lancet website so it's a bit hard to judge how much of this is true. The Guardian is hardly impartial in the matter.

It will be interesting to see how the methodology stacks up, but any scrutiny will not be until after the election.

This article has a bit more info. Apparently the study also claims that there was an increase in births from 275 to 366 over the same period. That's a 75% increase in the birth rate. Which seems unlikely.

But if true and if the argument is that these population changes have been brought about by the war then it's hard to see just how the war would lead to any increase in the birth rate, let alone a 75% increase. So presumable if the figures are right, a big if, there must be something else going on.

Another reason for skepticism is that the Iraqi Body Count people estimate a far lower number. And if anything their figures would be on the inflated side.

The Independent reports the study claims that war related violent deaths only occurred in 15 of the 33 neighborhoods sampled. Which doesn't make much sense if war related deaths were so much more wide spread than was thought. Presumably the study is extrapolating from the 15 to Iraq as whole, but given that that 15 is not even representative of their sample then this sounds risky.

But the articles may not be giving the study its due, I'll wait until I read the actual thing before making a judgment. If true it undermines the pro-war case.

UPDATE: the online report is here

Further update: having read thru the study I have some doubts but I'm not an expert in these sorts of studies so I'll wait till the experts have gone thru it. I note that the study chooses to exclude Falluja since they believe the data to be unreliable. If included the total would rise to 200,000. They have reasoned that there is a high likelihood of sample bias in this case.

ANTHER UPDATE: the exclusion of the Falluja data is interesting. It's excluded because the authors believe there is selection bias. But what if there isn't? That would imply that Falluja is an outlier because that is where very heavy fighting has occurred.

Presumably fighting did not occur evenly throughout Iraq. But the study assumes that to be the case - the mortality statistics apply to all communities whether or not fighting took place there. But if one assumes that fighting was clustered then the study's extrapolation would be on the high side. Especially as the authors believe that most of the war related deaths were from bombing. The bombing campaign was highly targeted so one would expect high casualty rates in those areas of bombing and far less outside of those areas. The areas with high mortality may not be representative of the country at large, just of those where fighting occurred. This is suggested by the fact that 15 of 33 communities sampled had all the deaths.

If this is the case then the study's data should be extrapolated out to those populations that experienced bombing, not the entire population.

On second thoughts that's not quite right. I think it just implies an increased sensitivity to cluster bias. If the 53 deaths at Falluja can cause the estimate to jump by 100,000 then presumably the same distortion can and may have occurred with regions of high fighting that remain within the data.

One would need to know the geographical spread of the 15 communities with fatalities. If many of them cluster around relatively small population centres and/or areas of the heaviest fighting then this could lead to over estimation.

5 Comments:

Blogger Greyshade said...

It's a big extrapolation but it still lookslike alot of excess deaths. The periods before and after the war are unequal (15 months versus 18 months) so you have to scale before you compare. On that basis 275 births before the war would correspond to 330 after the war so that the increase is only about 10% - it is statistically significant but could be explained by "household consolidation" (eg soldier killed so his wife and children go back to her parents house which is in survey but they don't report births or deaths from before they went back). Since there was only 1 violent death before study virtually all the reported violent deaths can be attributed to the war. Here the crucial thing is whether you leave out the Fallujah data - if you do you seem to get an excess of about 20 deaths from violence which with a multiplier of 3000 (total population 24m versus ~8000 for sample). Non-violent deaths seem to show an excess of about 15 and should be less patchy. I suspect that an excess of 40-50,000 non-violent deaths attributable to the war (sickness, malnutrition, etc) is quite plausible but would question whether there were 60,000 violent deaths given that body count specifies only about 15,000.

29 October 2004 at 5:32 PM  
Blogger Sock Thief said...

It's very hard to read. It may turn out to be true but in the end the study comes down to interpreting the meaning of 23 deaths (excluding the Falluja data). It will be interesting to see what comes out of peer review.

29 October 2004 at 5:56 PM  
Anonymous Anonymous said...

War causes death just like sanctions cause death that is unavoidable.

Anyone who thinks the war was going to reduce "excess deaths" (a particular accounting mechanism) directly afterwards was kidding themselves. But at the same time there are SOME occasions when one must make those sacrifices.

Was Iraq one of those?
I don't know.

For example it might have been better to let it blow out of control (for the US to be isolationist from the beginning) and see if the UN could deal with it without US help. But what I feared the most (and said at the time) was a "defeat" for those who care and a loss of will-power to EVER take action the same sort of loss of will power we had before WWII.

29 October 2004 at 9:36 PM  
Blogger Chef said...

Other studies of pre-war Iraqi mortality produced very different numbers than the Roberts et al study in Lancet. When you compare the various studies, I think it may be that the Iraq War did reduce "excess deaths" of Iraqis.

Certainly, the fact that Roberts et al come up with such different number than the other studies is something they should have discussed, but didn't.

See my blog post on the topic.

30 October 2004 at 8:03 PM  
Blogger memphis-slammer said...

Hype blog. And I admire your site and plan on
returning to it! When I web surf it always helps me to
find great blogs.
Once you sign on, check for my united cash advance blog.

18 January 2006 at 7:49 PM  

Post a Comment

<< Home