You would think that after all this time, all possible erroneous arguments against the Lancet study would have been made, but folks keep coming up with new ones. R.J. Rummel has come up with some new ones. Unlike many of the critics, Rummel has read the study; but unfortunately he has badly misunderstood it. Rummel writes:
The pre-invasion statistics were compiled by Saddam's Ministry of Health. There are questions one must ask of such a source that the Lancet researchers do inadequately. Did the ministry include murders or massacres by the Iraqi regime, such as in prison. To what degree did it include the deaths of children from the lack of food and medicine due to Hussein's refusal to use for that purpose the funds the UN set aside from his oil sales. Moreover, the way pre-invasion deaths were defined excludes periods of mass Iraqi deaths, as in the war Hussein launched against Iran in which overall a million were killed, and the mass murder of Kurds in the north and Shiites in the south to repress their rebellions. As a rule of thumb, never accept surveys or statistics produced under a bloody tyranny---what people say or do under such a system is out of fear.
The pre-invasion statistics were not compiled by Saddam's Ministry of Health. The article is perfectly clear on this. The summary says:
Methods: A cluster sample survey was undertaken throughout Iraq during September, 2004. 33 clusters of 30 households each were interviewed about household composition, births, and deaths since January, 2002. In those households reporting deaths, the date, cause, and circumstances of violent deaths were recorded. We assessed the relative risk of death associated with the 2003 invasion and occupation by comparing mortality in the 17.8 months after the invasion with the 14.6-month period preceding it.
So it includes murders by the Iraqi regime and any deaths of children because of shortages of food and medicine. It does exclude deaths in the Iran war and rebellions. The survey was not conducted under a bloody tyranny.
Rummel goes on to claim:
No tests of significance were applied to any of Lancet's statistical results. Why?
This is not true. They frequently give 95% confidence intervals for their results.
Rummel then takes Iraqi health ministry figures showing 3,274 deaths in military and terrorist conflicts in six months and multiplies
by 3 to get comparable time periods, which would mean about 9,822 civilians killed by comparison to lancet's estimate of over 100,000; 38 percent due to the terrorists versus 4 percent for Lancet. Hmmmm.
This is, of course, comparing apples with oranges. The Lancet estimate of 100,000 is of excess deaths. As well as deaths in the conflict it includes the increase in murder, accidents and disease that followed the invasion. Furthermore, the health ministry numbers are guaranteed to be an underestimate, since not every death will be recorded by Iraqi hospitals.
And Victor S takes a chainsaw to another critic of the Lancet report. As he writes,
if you're sick of repeatedly arguing with the bookless, innumerate cohort who reject the findings of the Lancet report on post-invasion mortality in Iraq, turn away now.
Here's a new (to me at any rate) criticism of the study. Deaths are seasonal. In the west more people die in winter than in summer, in the tropics deaths are still seasonal though seasons are different.
Now the Lancet study compared the 17.8 months after the invasion with the 14.6-month period preceding it. As we know mortality series have a 12 month cycle having variable "windows" will under or estimate mortality depending on whether peaks or troughs are double sampled. If you'd done this study in London and double counted Sep-Feb post-event and Sep-Oct pre-event you'd (substantially) over-estimate mortality even if no real change had taken place.
Any idea of what the Middle East mortality pattern is how the Iraq sampling window fits it? It might be worth looking at.