Iraq death toll: a response
Josh Dougherty, Researcher and Analyst at Iraq Body Count, has emailed me with some comments about my post yesterday on the PLoS study of the Iraq death toll. I thought his response was interesting, and reproduce it in full, with his permission:
A couple further points to note. The raw data has been made available on the PLoS website. That’s actually one of several improvements made over the prior Lancet report from 2006: in this case, more transparency. So it’s possible to crunch the numbers a little further.
I think the 95% interval for just violence will probably be narrower than you suggest, still wide but not proportional to the interval on the excess deaths.
I also think the estimate for war-related violence from this data would be a good bit lower than the 250,000 you suggest. The paper’s claim that “more than 60% of excess deaths were directly attributable to violence”, is I think kind of slippery. It seems like “violence” there must also be including what they call elsewhere “injury (not conflict related)” (see Figure 1). Or, more specifically, including the “excess” of that category as compared to the 2001-02 rates of the same category. So that seems to be what pushes it up to “more than 60%” of the excess death estimate. And when they say “excess deaths… directly attributable to violence”, this apparently means all of the “injury (conflict related)” plus the portion of “injury (not conflict related)” that falls above the pre-invasion rates of “injury (not conflict related)” measured for 2001-02.
It has to be including something beyond the violent war deaths to get it to be “more than 60%” of 405,000. For example, this new study and the Lancet one from 2006 cover almost exactly the same number of households. This study finds 76 deaths from war violence. Lancet found 300 violent deaths in the 2006 study. Those 300 deaths equated to “600,000″. Therefore, 76 deaths from war violence can’t equate to 250,000 in this new study. In order to get that high you have to include quite a bit of something else. In this case, I think that’s the ‘excess’ number of the “injury (not conflict related)” deaths. But again, they’re being slippery in a lot of ways with violence. That they don’t give a clear estimate, or a 95% interval, for violent deaths, despite repeatedly giving figures about violence trends and subsets, and implying a vague total with things like the “more than 60%” sentence, is, odd, to say the least. A skeptic might think they chose to hide such an estimate because it could then be easily compared to the estimate given for violence in the 2006 Lancet, and the obvious (and inconvenient) conclusions would follow.
I’m not sure what exact kinds of deaths fall into the “injury (not conflict related)” category here either. I don’t think there’s any clear explanation of this in the paper. Vehicle accidents would be one guess, but it may be a variety of things. Without knowing how they’re defining these categories it’s hard to tell exactly how the claim about “violence” being “more than 60%” of 405,000 actually relates to other things such as Iraq Body Count or Iraqi Family Health Survey, since it’s not really clear what they mean when they say “violence” there.
In any case, the corresponding number for Iraq Body Count for 2003-11, with combatants included as they are in the survey, would be around 160,000. So the findings on violent war deaths here would be somewhat higher, but not by a huge amount, and look pretty reasonable to me. The trends over time and proportions and other aspects they discuss about the violence also seem pretty reasonable, allowing for the wide margin of error around most of these things.
For some unclear reason again, they do give a specific estimate for war-related violence from the Sibling Survey of “132,000 (95% UI 89,000–174,000)”, covering persons aged 15-60, which would account for the vast majority of violent deaths in the war. This survey actually comes out lower than Iraq Body Count. The estimate they derive from this survey for excess deaths also comes out far lower than the one from the Household survey. Note as well that they specifically say “war-related violence” when giving this Sibling estimate, but don’t speak of “war-related violence” when giving the “more than 60%” claim for the Household survey estimate.
If any of the PLoS authors would care to respond, I would be happy to post their responses here too.Tagged in: iraq body count
Latest from Independent journalists on Twitter