Did President Bush Actually get Something Right About the Death Toll in Iraq?

Last week he was asked about a new estimate of the number of Iraqis killed as a result of the U.S.-led invasion. He said the paper was "pretty well discredited."

The study, published in the British medical journal The Lancet, gives a range of 400,000-800,000 deaths due to the war and its violent aftermath. That's not only way out of line with U.S. government estimates -- some ten times less -- but also with other counts. Iraq Body Count, a London-based antiwar group that is no way a shill for the occupation, puts the number of civilians killed since 2003 at 49,000. The United Nations Development Program estimate is 18,000-29,000 up until mid-2004.

People who oppose this misguided war will be tempted to accept the larger number offered by the Lancet authors, Gilbert Burnham, Riyadh Lafta, Shannon Doocy and Les Roberts. Obviously, though, their numbers are so different because they use a different method of getting their information. Unless you value propaganda above truth, then, you need to know: Is their approach more accurate than that used by others?

The Lancet team sampled 1,849 households all over Iraq, counted incidents of violent death in each, before and after the invasion, and then proposed that those rates apply to the entire population of 27 million Iraqis. This is how epidemiologists would estimate rates of death from measles or malaria -- take a sample and then project it onto the whole. The assumption there, of course, is that all people exposed to an epidemic are equally likely to catch the disease.

Is that assumption safe? Is death in war a phenomenon like measles? Neil Johnson, a physicist at Oxford who studies patterns in violent conflicts, believes that the answer is no. In an epidemic, he told me, it is safe to assume that all people have an equal chance of catching the disease. But in war and its violent aftermath, this assumption can't possibly be correct. Differences among Shiites and Sunni, politically engaged and remote, rich and poor, region and region -- all of these seem likely to have an impact. So if epidemiological methods produce a number out of line with other methods, it is at least worth asking if the problem is with the epidemiologists.

Johnson and his collaborator Michael Spagat, an economist at the University of London, Royal Holloway, believe they have found just such a fatal flaw -- an error that stems from mistakenly assuming that war deaths work like deaths from illness. It seems all the households surveyed by the Lancet authors were on main roads or at intersections of smaller streets with major arteries. It does not take a detailed knowledge of warfare to imagine how households in those locations would be more likely to report car bombings, market-place explosions, and attacks on vehicles.

For more on the burgeoning controversy, see today's reports in Science, and Nature. And keep in mind that the issue is not which political stance is better served by these figures, but simply whether they are sound.