The Los Angeles Times reported Friday on a poll from Opinion Research Business, a British polling agency. The poll suggests that more than a million Iraqis have died from the conflict resulting from the U.S. invasion and occupation.
The poll asked people how many in their household -- people living under the same roof - had been killed due to violent conflict since the U.S. invasion. 22% said at least one member of their household had been killed. Based on their responses, and the estimated number of households in Iraq, they estimate a total of 1,220,580 deaths since the U.S. invasion in 2003.
The Times reports:
It was the highest estimate given so far of civilian deaths in Iraq. Last year, a study in the medical journal Lancet put the number at 654,965, which Iraq's government has dismissed as "ridiculous."
It's important to recognize that the results of the ORB poll are quite consistent with the Lancet results, when you take into account the fact that the ORB poll was completed a month ago, whereas the Lancet cluster survey, the results of which were published in the fall, was completed in July 2006. A lot of people were killed in Iraq since July 2006.
Just Foreign Policy created an online and ongoing estimate that extrapolates from the Lancet study using the trend implied by the Iraq Body Count tally of deaths reported in the Western media. Our counter today stands at 1,044,607.
One can't compare the two numbers directly in a strictly statistically rigorous sense, because the two methods don't ask exactly the same question (the ORB poll relies on self-reporting of "death due to the conflict" whereas the Lancet study compared estimated violent death rates before and after the invasion), the ORB reported margin of error of 2.4% is a margin of error on the responses, not on the death estimate, and the Just Foreign Policy estimate is an extrapolation from the Lancet estimate.
Nonetheless, rough calculations show the two numbers to be basically the same.
Suppose that the true percentage of households in Iraq experiencing no deaths due to conflict since the invasion were one margin of error higher than in the ORB sample, so that rather than being 77.8%, it was 80.2%, and that the other responses (those reporting violent deaths) were scaled down proportionately. The effect on the ORB estimate would be to reduce it to 1,088,625 - roughly 40,000, or 4% higher than the Just Foreign Policy estimate. If the true percentage of households experiencing no deaths due to conflict were two margins of error higher, the effect would be to reduce the ORB estimate below the Just Foreign Policy estimate.
Or, coming from the other direction, consider the 95% confidence interval for excess violent deaths reported in the Lancet study. [This is the baseline for the Just Foreign Policy extrapolation, not the more widely cited number of excess deaths.]
The estimate was 601,000. The 95% confidence interval was (426,000, 794,000.) If we had used the right-hand endpoint of 794,000 instead of the estimate of 601,000, the Just Foreign Policy extrapolation would yield 1,379,404.
In other words, roughly speaking, the ORB interval covers the extrapolated Lancet estimate and the extrapolated Lancet interval covers the ORB estimate. To the level of precision that we know these numbers, and taking into account the key difference that an increasingly violent year transpired between the Lancet estimate and the ORB estimate, the two estimates are the same.
Until now, many -- including critics of the war -- have shied away from the Lancet estimate because of the lack of independent confirmation. Unfortunately, this has led many to cite the Iraq Body Count tally of deaths reported in Western media as if it were an estimate of the death toll, which it is not. Now that the order of magnitude of the death toll reported by the Lancet study has been independently confirmed, pressure should be redoubled on media outlets to tell the truth about the Iraqi death toll. As Congress is currently debating efforts to end the war, there could not be a more appropriate time to do so.