Let me preface this diary by saying the recent Johns Hopkins study on mortality in Iraq may well be accurate, and that I'm not debating the larger questions of US complicity in the war or culpability in Iraqi deaths. I agree with the basic rationale for the study, which is that mortality rates based on morgue information and public news sources underreport violent deaths. However, I was a little shocked by the findings of this report. If the report is true, the death rate from this current war is equal in scale to the Iran-Iraq war, which lasted a couple years longer, but cost an estimated one million lives.
Few Iraqis were untouched by that war, and nearly every family I know had someone killed or injured. The Hopkins report indicates that one in fifty Iraqis died violently so far in this war. However, of the more than 300 Iraqi health care workers we train, supervise and monitor in ten governorates, none have been killed or injured, and to my knowledge, only one has been attacked. She escaped a kidnapping attempt in Baghdad. No relatives of our staff have died violently or been injured. Thank God for that, and may our luck continue. But the scars of the Iran-Iraq war seem much deeper to me than the current war, despite its immediacy. And yes, we do work in some of the hot governorates, including Ninevah, Diyala and at-Tammim.
Here are some possible weaknesses in the John Hopkins study, based on the PDF Lancet article The Human Cost of the War in Iraq. The article can be found here: http://i.a.cnn.net/...
First, the most violent governorates are relatively oversampled. The provinces experiencing full-scale war - Anbar, Ninevah, Salahaddin, Diyala - (>10 violent deaths per 1,000 per year) were sampled at a rate of one cluster per 459,000 people. Baghdad was sampled at 1:540,000. In predominantly Shia' governorates that have experienced some inter-shia' political violence and some bombing incidents (Babel, Qadisiyya, Basra, the rate is one cluster per 809,000. In Kirkuk (Tamim), where violence is highly variable but with large areas that are peaceful, the rate is 1:881,000. In areas without significant violence, one cluster per 530,000 was sampled - but they did not survey the two most secure governorates in the north and south - Dohuk and Muthanna, respectively. Dohuk is the only Kurdish governorate that has experienced no fighting and no bombings of any sort. Likewise, Muthanna is the calmist governorate in the Shia' area. It's so calm, the US sent the Japanese there. Admittedly, these are small governorates, but they do have an aggregate population of 1.5 million people who are essentially unaffected by the war, other than soldiers recruited there who agree to fight elsewhere.
Put another way, the Sunni governorates were sampled 1:450,000; the mixed ethnicity governorates sampled 1:532,000; the Kurdish governorates 1:626,000 and the Shia' governorates 1:660,000. Violence is far higher in the Sunni and mixed-ethnicity governorates, because the fight between the US and the insurgency is in primarily Sunni areas, and the civil war is primarily in mixed ethnicity areas. Violence is lower in Shia' areas and very low in Kurdish areas. Finally, the populations in the Sunni and mixed ethnicity governorates may be slightly overestimated for two reasons: First, the UNDP data is based at least partially on Iraqi census figures before 2003, which tended to undercount Shia' and Kurds, and second, there has been massive migration out of Baghdad, Ninevah and Diyala governorates to safer, ethnically homogenous areas since the war - there are 250,000 registered IDPs in Iraq, but there could be twice that many or more who have quietly moved in with relatives outside of the most violent governorates.
My biggest concern however, is that violence is highly unequally distributed within governorates, both geographically and according to ethnic community. If there appears to be an unintentional sampling bias toward the most violent governorates, there could also be a trend to sample the more violent locations within each governorate. I know the report states that clusters were selected randomly, but the locations of those clusters are really important for assessing accuracy. For example, the study only sampled one cluster in Kirkuk (Tammim). If you survey a mixed-ethnicity neighborhood near the center of the city, the mortality rate would be sky high, among the highest in Iraq. If you measured an ethnically homogenous neighborhood in the city, the rate would be moderate to high. If you measured an ethnically homogenous village west or south of Kirkuk, the rate would be very variable from relatively high to low. If you measure a town or village in the east of the province, the rate would be negligible. It seems to me very hard to get an accurate reading on Kirkuk from one cluster.
Likewise, they used three cluster sites to determine the mortality rate for Ninevah governorate, Iraq's second largest governorate. The northeastern third and about 35% of the population are under Kurdish control and experience virtually no violence, rural areas and areas along the Syrian border experience localized violence depending to a great extent on the ethnic composition of the community, and Mosul city is insanely violent. The location of those three clusters is really important, even within Mosul city itself. The west side of town is twice as violent as the east. Without information on the location of the clusters, it is hard to be 100% convinced of accuracy. Diyala is similar - with extraordinarily violent areas (Khalis, Baquba) and relatively safe ones (Khanaqin, Kifri). I can travel safely to Khanaqin and have lunch in a restaurant, but I would be immediately killed or kidapped if I tried that in Baquba.
Unfortunately, the ethnic affiliation of the surveyor is also important (i.e. Arab communities would not accept a Kurd and vice versa). I know that they achieved gender balance, but it is hard to imagine how one could get accurate figures in mixed ethnicity governorates like Diyala or Kirkuk without first, a number of clusters and second, withou careful attention to assure an ethnic mix of researchers to assure trust on the part of participants and accurate interviews. They may well have done the latter, but it is not stated in the report.
I offer these critiques as grains of salt. The report may in fact be accurate. I do not dispute the honesty of the researchers. I know from experience that one never has much control over operations in Iraq, and without a great deal of control, information errors can creep in. Despite what appears to be oversampling of the most violent parts of the country, the Hopkins study may yet be accurate if sampling clusters within governorates do not skew to the more violent neighborhoods or districts.
My own guess is that the death rate in the war is twice as much or more than Iraq Body Count, but probably half as much as reported in this study. That's still far too much violence and too many deaths. My comments are not intended to shift the blame from Bush or the incredible failures of the Administration to secure the country and transition to Iraqi rule. I just want to be a little careful, because of the risk that any inaccuracies will be used against us. One of the authors was a primary candidate for Congress in NY24 (Les Roberts) and that fact will immediately draw political heat from the right. That noise may obscure the fact that even a death rate half of what the Hopkins report suggests is actually far higher than the Administration reports.
Accuracy and caution is also warranted because we need to watch carefully the rate of ethnic and sectarian violence in Iraq. Sunni-Shia' violence threatens to increase massively above current levels, and the US needs to have some sort of a plan rather than the stream of platitudes, wishful thinking and good news propaganda coming out of this Administration. Policy people on our side of the isle need to be accurate, sober and be ready to come up with some alternatives based on the shift in violence toward open sectarian civil war and possibly genocide. When a real genocide starts developing, we want to spot it.