Iraq War Dead Estimate Controversy

You might find this interesting, as I did — a question from a friendly Midwest blogger about the Lancet study (PDF) using statistics to estimate the number of Iraqis killed during our intervention in Iraq. Criticisms vs. the study were raised by an article in Slate, and those criticisms were (I think) well answered by an article in Crooked Timber. First the question I got (with the links); then my response…

===============================

I was discussing this article from Slate by Kaplan who disagreed with the methodology.

However, someone responded with a link that argued that the criticisms weren’t valid.

From what I can understand, the dispute was centering around whether the regions were representative and whether the 95% confidence margin was too wide to invalidate the poll. They seem to be in dispute about what even the key “rules” are to determining this sort of thing so I know I’m out of my league arbitering it.

=====================================

OK, I’ve had a chance to read both articles. When I was reading the first, I was making mental notes; and lo and behold, the second author (in Crooked Timber) made the points I was thinking of, and several others and better than I could.

The first author raises a good point about the wideness of the confidence intervals; the second author correctly points out that the most likely value in a confidence interval is the central value — the likelihood of other values in the range drops off dramatically the farther away you get from the central value (giving rise to the characteristic “bell curve” describing a “normally” distributed population — basically any large population). It’s just much more likely in an honest sampling to get a value, say, 1 point away from the average than 2 points away; much more likely to get a value 2 points away than 3 points away; and so forth.

You can conduct an experiment yourself: Try flipping a coin ten times. You’ll probably get heads five times. You’re less likely to get heads four or six times, and you’re a lot less likely to get heads three or seven times — try repeating your ten-flip experiment several times; I bet those are the results you get (of course, you could just get “lucky” and get heads every single time; but that’s highly unlikely, by common sense and statistics, describing the real world).

I might only add that the wideness of the confidence interval should give us more, not less, confidence in the researchers: They are erring on the side of caution, not recklessness, by reporting such a wide range.

The sampling method seems to me (not an expert on stats, just a student of the fascinating subject) to be reasonable, for all the reasons the second author cited. His minefield analogy was brilliant: You can see how such “cluster” sampling (sampling randomly selected neighborhoods nationwide rather than, say, every tenth person all across the country — an impractical alternative to this cluster method, a historically reliable method) will tend to UNDERestimate, not overestimate, such “rare” events as violent death — you’re “lucky” to hit any pockets of such unusual occurrences (assuming that violent deaths are “unusual” in the population — even during wartime most people don’t get killed, which is why we’re looking at these estimates).

Personally, I think the authors of the study should have included, not excluded, the data from the neighborhood that their random sampling happened to select in Falluja, even though it showed a much higher death rate than elsewhere in the country: Of course it did, because it was a center of resistance; and it was not the only place in the Sunni Triangle that was so inclined and so hard hit — that’s the whole point of cluster sampling over the whole country (the samples selected by random global positioning points) — if you’re in a “minefield”, you’re going to hit a “mine” every so often (to use the second author’s analogy).

The comparison by the first author of death rates in a pre-war study (whose possible and perhaps probable sources of error are examined by the second author) with pre-war estimates in the second study is not terribly useful: The two studies were very much unlike, in methodology and focus (ex. infants and adults), so it’s rather like comparing apples and oranges (If the Lancet study had used such comparisons, its critics would really be having a field day). What’s more the issue is how significantly death rates have increased, using the same methodology and samples etc. in the Lancet’s pre- and post-war analyses. And as the second author points out — and as common sense dictates — the death rates in Iraq have evidently increased significantly and probably dramatically with the war and its aftermath.

The issue of “lying Iraqis” seems the strongest to me; however, the second author points out that in at least 81% of the cases in which the veracity was challenged, there were death certificates available to back up the reports of deaths (One could assume all sorts of conspiracies from there onward, but there is apparently no evidence to contradict this finding). The data appears legitimate.

And the second author does well in quickly debunking the comparison of the Lancet findings with the confirmed death totals reported to the media; obviously, in wartime, a lot of deaths go unreported — again why we have to try to estimate deaths from the best samples we can get.

How many people — civilians, combatants, civilian/combatants — have died in Iraq because of the war? No one really knows or ever will know with a great deal of precision — that’s the nature of warfare, down through the ages. Regardless of how it has been over-hyped or legitimately or incorrectly critiqued, this study does seem to make a serious atttempt, using legitimate and time-tested methods, to give us some sort of idea of the real human cost of our intervention in Iraq, most likely measured in the tens of thousands of lives.

I must add, however, that if we are to make a full accounting that we must add in the deaths (not to mention maimings etc.) resulting from the additional interventions our government has made in Iraq over the years (primarily to maintain a stable supply of oil): Our government helped install and maintain Saddam Hussein, who used murder as a tool not only to maintain power once he got in but also to gain power in the first place, which was well known to our government at the time; our government backed Hussein in the Iran/Iraq War, which he started and which resulted in the deaths of some one million people; our government backed Hussein when he gassed the Kurds (An infamous, contemporaneous photo even shows Rumsfeld shaking Saddam’s hand; and components and/or delivery systems for the poison gas were, I believe, supplied at least in part by American companies, under the supervision of our government); and our government encouraged the Shiites in the south to revolt after the first Persian Gulf War, then stood by and allowed Hussein to slaughter them (resulting in those mass graves we’ve heard so much about).

We’ve had the blood of hundreds of thousands of Iraqis on our hands for decades. Ironically, by getting rid of Saddam (who no longer was a cooperative puppet), we are “cleaning up a mess” that we had a big hand in creating. Hopefully, we will make things better for the common Iraqi people in the long-run; but I doubt they will have a government that will be much less of a cooperative puppet than that of Saddam Hussein pre-Gulf War I.

As Ronald Reagan so bluntly put it, our interest in the Middle East can be spelled in three little letters: “O – I – L”.

We ain’t invadin’ the dictatorial republic of No-oil-istan.

Today, Iraq; tomorrow, Iran. What’s a consonant between friends?

-Doug D.

Leave a Reply

Your email address will not be published. Required fields are marked *

Connect with Facebook