Friday, May 27, 2005

Expert Witness

I am pleased to see that Debarati Guha-Sapir and her colleagues at the Center for Research on the Epidemiology of Disasters (CRED) have made public an analysis of mortality in Darfur (pdf). (Nod to Coalition for Darfur, naturally. There have been many posts there over the last two weeks worthy of discussion, but since I made a point to criticize Guha-Sapir, I couldn't let this one pass by unremarked.)

Guha-Sapir et al. estimate the range of excess deaths between September 2003 and January 2005 to be 120,000 according to what they call the CRED method, or, using the State Department method, the estimate for excess deaths between March 2003 and January 2005 is either 63,000 or 140,000, depending upon whether one chooses a low or a high estimate of crude mortality rate (cmr). It's my opinion that the lowest reported cmr estimates have tended to come from small, accessible, secure, well-managed camps, whereas the larger cmr estimates have tended to come from very large camps with persistent security issues. I cannot regard the latter as unrepresentative because they in fact account for more people, although extrapolations based on such numbers are problematical.

The CRED estimate relies upon a number of assumptions and a certain interpretation of events that may be called into question. By far the greatest factor accounting for the discrepancy between the CRED estimate and the estimates provided by the Coalition for International Justice et al. concerns the question of whether the WHO data have been misunderstood:

The WHO mortality survey and the WHO mortality projections have often been confused and misguidedly used interchangeably. This has led some to misinterpret a WHO statement indicating exclusion of violent death from the WHO estimate, as also meaning violent deaths were not included in the WHO mortality surveys. This wrong assumption has led to double counting of violent deaths in many subsequent projections.

I tend to agree with Guha-Sapir et al. on this point, although Eric Reeves has indicated that this is not the whole story, and that a clarification from the WHO would settle the matter of whether the survey reported just some or all violent deaths.

I disagree with Guha-Sapir et al. on other key matters, such as the nature of the data from the Atrocities Documentation Team (ADT), the progression of events in Sudan which is crucial to their disaggregation of data, and how best to characterize the overall situation--pockets of high mortality that contain hundreds of thousands of people tell us more about the overall situation than pockets of stability that contain far fewer people. Therefore I am inclined to regard the CRED estimate as rather low, and the low estimate of the State Department as not particularly credible. However, as the CRED estimate appears to be rather credible at this stage--it is certainly well argued--and as it is significantly lower than my rough sense of how many people have died, I will have to reevaluate my views.

I am grateful to Guha-Sapir and her team for giving this matter the serious attention that it deserves.

posted by Fido the Yak at 10:37 PM.


Anonymous Olivier Degomme said...

Hello Fido the Yak,

As one of the authors of the report you are citing ("Darfur : Counting the Deaths"), I'd like to thank you for your interest in the subject. It also seems from your comments that you gave the report a thorough look, which pleases me a lot. However, I'd like to have your inputs on the different items you disagree with.
Looking forward to hearing from you.

June 03, 2005 3:27 AM  
Blogger Fido the Yak said...

Hello, Dr. Degomme.

I suppose I can expand on my criticisms.

Regarding the narrative of events, I find it suspect on a number of counts. The report cites a State Department document that I believe Eric Reeves has thoroughly lambasted for, among other things, its lack of citations to its sources. The lack of citations isn't my main concern, though it's perhaps a key to the issue. It seems pretty clear that the narrative is not emerging from the epidemiological data. The periods demarcated do not neatly correspond to the temporal fluctuations that may be seen in various the epidemiological surveys. This is my view, which is obviously debateable. It wouldn't be a matter of debate if the narrative episodes were derived solely from the mortality data. The report argues that a narrative of events is needed to correct a distortion that would appear should all of the data be treated synchronically. That may well be true. However, it does not follow that a narrative structure must be imported wholesale, or that any recounting of events must be taken as given. On the contrary, I would think one has a hermeneutic responsibility, as it were, to explore and explicate if necessary the meanings of key events. What, for instance, does one make of the term "cease fire" in the context of the situation in Darfur? What meanings does that term have for the various agents most probably responsible for the rather exceptional mortality data under consideration? This is not an ancillary consideration in my view, but goes to the heart of the problem. I and many others I should think look to mortality data as evidence of the nature of the violence that has been occurring in Sudan. In reading epidemiological reports to this end, then, it's important that I understand not only how the violence is being characterized, but the reasoning behind the characterizations, and their relationship to the data and the generalizations made from the data. I did not get that sense from reading "Darfur: Counting the Deaths." It struck me that other configurations of the data might be equally intelligible, and it occured to me that the variation and the number of samples might be such so as to preclude any reasonable attempt to induce a coherent narrative from the raw epidemiological data.

The other point that especially bothered me was the dismissal of the evidence from the Atrocities Documentation Team primarily on methodological grounds, especially as the evidence contradicted one of the report's major arguments concerning the nature of the violence. Well, that is an argument. Your team presented its case, but I was not persuaded. I don't know what more I can say.

September 20, 2005 8:20 AM  

Post a Comment

Fido the Yak front page