Very interesting.
[Jim Lindgren, May 15, 2005 at 2:01am] 1 Trackbacks / Possibly More Trackbacks
Are there 24,000 extra Iraq War deaths or 98,000?–
- These Different Estimates Can Be Reconciled.
There is a dispute over how to reconcile the recent published estimate of 24,000 extra war-related deaths in Iraq (Iraq Living Conditions Survey 2004 Iraq | United Nations Development Programme ) with the previous Lancet study, which estimated about 98,000 extra deaths from the Iraq War ( http://www.zmag.org/lancet.pdf ).
Tim Lambert has argued ( http://elgar.cse.unsw.edu.au/~lambert/cgi-bin/blog/2005/05#lancet34 ) that these numbers are really fairly close when you recognize that the 24,000 figure represents only war-related deaths, while the Lancet study involved extra deaths from all sources. He criticizes the reporting of the Times of London:
Unfortunately, the Times reports the ILCS results like this:
[i]The 370-page report said that it was 95 per cent confident that the toll during the war and the first year of occupation was 24,000, but could have been between 18,000 and 29,000. About 12 per cent of those were under 18.
The figure is far lower than the 98,000 deaths estimated in The Lancet last October, which said that it had interviewed nearly 1,000 households. But it is far higher than other figures. [/i]
This makes a misleading comparison between the Lancet number for all excess deaths (which includes the increase in murder, accidents and disease) and the ILCS number for deaths directly related to the war (which just includes deaths caused by the coalition and the insurgents). It also misses that the time periods were different.
The authors of the ILCS report (page 55 Iraq | United Nations Development Programme ) seem to have been confused themselves on this point, which makes the resulting confusion of commentators and journalists fully understandable (tip for the quote to Tim Blair http://timblair.net/ee/index.php/weblog/iraqs_dead_counted/):
[i]The ICLS data indicates 24,000 deaths, with a 95 per cent confidence interval from 18,000 to 29,000 deaths. The confidence level was estimated using a linearisation technique (using SPSS Complex Samples, version 12).
Another source (Roberts et al 2004) [Lancet] estimates the number to be 98,000, with a confidence interval of 8,000 to 194,000. The website ?Iraq Body Count? estimates that between 14,619 and 16,804 deaths have occurred between the beginning of 2003 and 7 December 2004.[/i]
I think that Tim Lambert is basically right (and he should be commended for sorting this out)( http://elgar.cse.unsw.edu.au/~lambert/cgi-bin/blog/2005/05#lancet34 ). The number of violent deaths in the Lancet report is not that much higher than the number of war-related deaths in the more recent ILCS study, and the period is slightly longer in the Lancet study. The 98,000 figure covers deaths from all sources (including accidents and disease), while the new ILCS study’s 24,000 estimate excludes deaths from non-War related sources of death, such as accidents or disease.
- Further Thoughts on Counting Deaths in Iraq.
But one should note that the 24,000 estimate includes all war-related deaths and disappearances of household members in the 24 months before being surveyed (most ILCS surveys were done in April and May 2004, a small % were done in August 2004). Thus, the 24,000 figure covers both most of the year before the March 2003 War and the year after. Any deaths of those in the Iraqi military before the War or anyone who disappeared in the year before the War would have been included in this 24,000 figure, if the household members treated these as War-related (though I doubt that deleting these pre-War deaths would cause large reductions in the estimates).
There is also the problem with both of these studies that people may be more likely to remember more recent events than earlier ones, and may be more likely to consider some people as a member of the household, people who lived with them recently more than people who lived with them two years ago (household size averaged about 7.5-8 persons in the Lancet study).
For example, the Lancet study distinguishes between the 14.6 month period before the War and the 17.8 months after the War. I find it somewhat odd that heart attack and stroke deaths are up 64% in the later period, and accidental deaths are up more than 3-fold. And live births are up 33% in the later (War & Post-War) period, even though post-War pregnancies would not lead to live births until 9 months had passed, so the rate of having children would likely have to have jumped substantially more than 33% in the last half of the later period. Further, household size jumps from 7.5 in the earlier period to 8.0 in the later period.
None of these changes is impossible, and some of these differences might be expected to some extent. But taken together, they give me pause before concluding that people are reporting deaths and who is a household member with the same reliability for before the War as for the post-War insurgency period, which is fresher in their minds.
Actually, there are some fancy statistical techniques to explore whether undercounts are more of a problem in one period than another, so-called capture/recapture techniques. If the researchers independently took the same information from multiple sources (such as more than one person in the same household interviewed separately on different days, or a next-door neighbor asked about their neighbor’s family and deaths, or a police or coroner?s list of deaths), the researchers could compare the various lists to see if they overlapped for the recent period more than for the earlier period. If they did, then the methodology would adjust the total death count more for undercounts in the earlier period than in the later period.
Personally, I wonder how much of the substantial differences between the earlier and later period (eg, number of births, household size, cardiac deaths, accidental deaths) are the result of differences in remembering events and household composition. But Tim Lambert is correct that the Lancet and ILCS studies tell roughly similar stories on violent death rates from the Iraq War itself and the post-War insurgency period.
One of the encouraging things about the blogosphere is that puzzles like the death estimate discrepancy can be sorted out.