Numbers games

So I read the Iraq war death report published in the Lancet earlier today, for those of us with subscriptions, Tim Lambert provides a link to the PDF version of the paper. Otherwise, here's the WaPo and the NYT versions.

What's striking is how the present study matches the 2004 study within that same timeframe. Here's other Iraqis being asked essentially the same questions--who lived here before the invasion, and who lived here every year since and why do they not live here anymore if they're gone--in a wide variety of settings throughout the country and they pretty much match up with the estimates of before, this time with a lower standard error. So if the Iraqis are lying, it'd be a pretty big conspiracy!

Another striking result is that violent deaths of all sorts have increased every year since the invasion; a fact that is also noted by the Department of Defense, and the Iraq body count--both are trusted by the conservative warhawks more than the Lancet. I spoke with someone who works for the Air Force yesterday and they just had a "convention" on IEDs in Iraq. He told me that the monthly IED count for the month of August there is somewhere around 2750 and steadily increasing. With so many poppers scattered about the streets, one could imagine a few of them going off on Iraqis--whether they are combatants or not.

To be sure, wingnuts are not happy. One often sited voice of dissent is Anthony Cordesman :

"They're almost certainly way too high," said Anthony Cordesman of the Center for Strategic & International Studies in Washington. He criticized the way the estimate was derived and noted that the results were released shortly before the Nov. 7 election.
"This is not analysis, this is politics," Cordesman said.

The article itself, however, points out possible scenarios how the numbers could be either too high or too low. As Mike points out, however, if you don't trust these numbers then you also shouldn't trust the numbers in Bosnia, the Congo and Darfur. In other words, if you use the same methodology in various scenarios of instability you can get a relative damage estimate for that instability compared to others. In that sense, the number itself is slightly less valuable than the relative value of that number, it could be 65,000 or 6.5 million as long as it's done the same way as other surveys were done you can compare that number to other disasters. And Iraq ranks right up there with all the major humanitarian disasters. Going to morgues and reading newspaper articles to come up with an estimate--like the Iraq body count does--usually underestimates the casualty figures by at least 50% on average, but if you're going to use those numbers, then compare apples to apples and use the equivalent statistics from the Congo or Nicaragua and you'll still come up with a large disaster, relatively speaking. Fact is, most people probably don't bring their dead to the morgue or identify them as dead to the authorities while hostilities are ongoing so direct surveys are much more powerful. And these researchers did everything to insure accuracy or results and randomness of the survey population. It's solid.

As for political maneuvering, puh-lease! Elections take place every two years so according to them, the only way for a scientific publication not to be politicized is for it to be published during off years. But when things like this are published during off-years, any number of legislative and executive functions are ongoing...so it could be perceived as politicized at a time when Bush's poll numbers are increasing, to distract from anti-terror legislation, etc., etc., etc. Data goes to the journal, the peer review process takes an indeterminate amount of time, then it gets edited for publication. Practically speaking, the only way this could be politicized is if the editors decided to bring it out on time or intentionally wait until after the elections. In which case they would have politicized science, just in a direction more comfortable to wingnuts. Incidentally, "the survey was done between May 20 and July 10, 2006."

Finally, yes: the study does not distinguish between "innocent" civilians and insurgents, terrorists, bad guys, whatever. This might actually need to be made more clear in the media accounts. It does not, however, negate the ability to use these data to compare to other humanitarian crises. Is this less or more bad than the Congo (Darfur, Bosnia, etc.) can still be asked when examining the data.

Update: more on death certificates in Iraq