2004 November 01 Monday
Latest Iraq Civilian Death Toll Estimate Too High?

Some folks at Johns Hopkins just rushed a report on Iraq death tolls into print in the British medical journal The Lancet so it would reach the public before the US election. Coverage by The New Scientist is typical in citing a very high and shocking estimate of Iraqi deaths as a consequence of the invasion.

The invasion of Iraq in March 2003 by coalition forces has lead to the death of at least 100,000 civilians, reveals the first scientific study to examine the issue. The majority of these deaths, which are in addition those normally expected from natural causes, illness and accidents, have been among women and children, finds the study, released early by The Lancet on Thursday.

Reuters offers similar coverage.

The rise in the death rate was mainly due to violence and much of it was caused by U.S. air strikes on towns and cities. "Making conservative assumptions, we think that about 100,000 excess deaths, or more have happened since the 2003 invasion of Iraq," said Les Roberts of the Johns Hopkins Bloomberg School of Public Health in a report published online by The Lancet medical journal.

But the Reuters report brings up the location of most of these estimated deaths: Fallujah.

Two-thirds of violent deaths in the study were reported in Falluja, the insurgent held city 50 km (32 miles) west of Baghdad which had been repeatedly hit by U.S. air strikes.

However, the researchers claim to have left out Fallujah's deaths when calculating their estimate for the total number of deaths.

"We were shocked at the magnitude but we're quite sure that the estimate of 100,000 is a conservative estimate," said Dr. Gilbert Burnham of the Johns Hopkins team. Dr. Burnham said the team excluded data about deaths in Falluja in making their estimate, because that city was the site of unusually intense violence.

So what to make of this report? Well, over on Slate Fred Kaplan bothered to look at the report and found the key sentence has a 95% confidence interval that ranges over more than an order of magnitude.

We estimate there were 98,000 extra deaths (95% CI 8000-194 000) during the post-war period.

That "CI" stands for Confidence Interval. There is a 95% chance that the real increase in casualties is between 8,000 and 194,000. That range is huge. There is nothing "conservative" about the 100,00 estimate. Kaplan goes on to say a great deal more about many problems with the study. Importantly, he argues that the 5 per 100,000 mortality rate that the researchers assume for Iraq before the war is much too low and the real mortality rate before the war might have been very close to the current mortality rate in Iraq. So this whole report's results may be built on one very unrealistic assumption.

At least the medical journal does acknowledge there are some serious limitations with this study.

Lancet Editor Richard Horton adds in an accompanying commentary: "The research we publish today was completed under the most testing of circumstances - an ongoing war. And therefore certain limitations were inevitable and need to be acknowledged right away. The number of population clusters chosen for sampling is small; the confidence intervals around the point estimates of mortality are wide; the Falluja cluster has an especially high mortality and so is atypical of the rest of the sample; and there is clearly the potential for recall bias among those interviewed. This remarkable piece of work represents the efforts of a courageous team of scientists. To have included more clusters would have improved the precision of their findings, but at an enormous and unacceptable risk to the team of interviewers who gathered the primary data. Despite these unusual challenges, the central observation - namely, that civilian mortality since the war has risen due to the effects of aerial weaponry - is convincing. This result requires an urgent political and military response if the confidence of ordinary Iraqis in the mostly American-British occupation is to be restored."

Will some of the Sunni Iraqis in particular tend to exaggerate the death tolls they experienced Also, how many of the civilian casualties happened during the initial invasion and how many have occurred during the occupation?

Read Fred Kaplan's full article and post your reactions in the comments here.

Share |      By Randall Parker at 2004 November 01 01:09 AM  Mideast Iraq


Comments
John S Bolton said at November 1, 2004 2:12 AM:

One might guess that the larger part of a death total in the tens of thousands would come from disruption, by war, of the medical capacities. Afterwards, there could be a lowering of the death rates to a level below that of the previous regime, as medicine improves.

AMac said at November 3, 2004 7:47 PM:

There are more controversial points to the Roberts et al. study than the single one covered by Slate's Kaplan. There are extended discussions at two web-logs. Shannon Love at ChicagoBoyz has three critical posts; dsquared ardently defends the study in her comments, and at his home blog, Crooked Timber.

John Ray said at November 7, 2004 10:12 PM:


http://dissectleft.blogspot.com/2004_10_31_dissectleft_archive.html#109943417654606465

$ said at February 25, 2005 10:14 AM:

i agree this estimate sounds unbelievably high but as far as the mortality rate goes... there's no way that it was the same as it is now

how could it be with bombs falling from the sky?

pcomeau said at July 12, 2005 5:40 PM:

Some statistical nitpicking...

C.I. is a measure of likelihood that the mean will be contained with in the data. In other words a C.I. of 95% means there is a 95% chance that the number x will be within two standard deviations of the mean. So with good sampling and a well run study the mean will be accurate to 95% (two standard deviations.)

The trick is to remember that this is being plotted on a bell curve that represents the probability of the number. The high and low are at the 2.5% part of the graph. So it is not a huge range per se, just where two standard deviations fall from the mean. This is common to statistics practice to make sure the mean falls within the densest area of the data.
(ref: Introduction to the Practice of Statistics Fourth Edition; Morre, David S. & McCabe, George P.; W.H. Freeman & Company; 2003; pp 417 - 428)

Sadly the any link to the study I've found requires payment, so I can't analyze the validity of the study from a point of view of soundness and good methodology.


Post a comment
Comments:
Name (not anon or anonymous):
Email Address:
URL:
Remember info?

      
 
Web parapundit.com
Go Read More Posts On ParaPundit
Site Traffic Info
The contents of this site are copyright