Why I got angry
Posted by Richard on October 18, 2006
I routinely have breakfast on Saturdays with a small group of libertarian friends, and those gatherings are generally enjoyable and intellectually stimulating. Occasionally, however, they’ve devolved into unpleasant arguments between David and me over Iraq, Gitmo, or something else related to the War Against Islamofascism.
Last Saturday, David brought up the new study published in Lancet claiming there’ve been 650,000 "excess" deaths in Iraq since March 2003 (I’ll post something about that study later). This led to an especially unpleasant argument during which I admit I became quite angry.
We’ve been friends for a long time, so we both made an effort to calm down and talk about it rationally. I had trouble explaining why I got so mad because I really wasn’t sure.
I’ve thought about it since, replaying the argument (as best as I can recall) in my mind. I think I’ve figured out what initially triggered my anger.
At one point, David said that when Bush was asked about the Lancet study, he said its methodology was flawed. David scoffed/sneered at the idea that someone who was "at best a C- student" and knew nothing about statistics would dare criticize the methodology of such learned scientists publishing in such a prestigious journal.
That moment, I believe, is when my blood really began to boil. Why? Was I angered at the insult to the Prez? Nope.
I think I get really mad at David when he deeply disappoints me. I’ve always considered him a thoughtful and intelligent person — someone whose thinking I respected even when we disagreed. But lately, on these topics, I’ve been hearing things that are unworthy of someone with his intellect. I expect better from him, and I get mad when he fails to live up to my expectations.
I became really angry at David last Saturday because he sounded exactly like Janeane Garofalo.
VRB said
Why would that be unworthy of his intellect. Many people including myself who graduated with better than a “C” average would not be capable of assessing that study. You may intuitvely think something is not rigth about the methodology, but you would not have any way to mathematically challenged it. I would say the same if you had several statistics courses. Statistics is something you would have to live and breathe to be able to accurately say what Bush said.
Anonymous said
Well, VRB, don’t you think that W is marginally familiar with the population of Iraq (about 26 million per Wikipedia?)
Don’t you think it’s possible he has enough mathematical capacity to figure out that 650,000 is over 2% of that?
I certainly don’t live and breathe statistics, but I did roll my eyes at the Lancet numbers. And without knowing anything about the methodology, it did occur to me a study that had almost 2 and a half percent of the population being killed off outright seemed likely to have a madness to its method.
Anyone heard of “common sense”?
VRB said
“You may intuitvely think something is not rigth about the methodology, but you would not have any way to mathematically challenged it.”
That is what I said. I did not make any judgement about the study. I can’t spell either.
Anonymous said
Rightly or wrongly, I didn’t interpret what David said (and how he said it) as a fair, objective statement that laypersons, Bush included, couldn’t adequately judge the study. It struck me instead as the same kind of contemptuous “Chimpy McBushitler is a moron” sentiment that you get from the Air America, DailyKos, moonbat left crowd — thus the Garofalo reference.
That said, it doesn’t take a statistician to see the problems with the study in Lancet — I think they’re obvious to a fair-minded lay-person. For just a few of the serious issues, check out what Iraq Body Count said. They’re a leftist anti-war group, and their assessment is friendly, polite, cautious — and yet utterly devasting.
I want to clarify something about my post, however. I was explaining my anger, not excusing it. I have no right to demand that David live up to my expectations.
David Bryant said
”I have no right to demand that David live up to my expectations.”
Maybe not. But as my friend, you certainly ought to let me know when I make a mistake. And I made a big one that day.
Just to be clear, Mr. Bush was quoted in the AP story as saying that the method of the ”Lancet” study “has been pretty thoroughly discredited.”
I’ve spent a lot of time this past week reviewing the ”Lancet” study and familiarizing myself with the available demographic data from Iraq. I still think it’s wrong to say that standard epidemiological methods have been “discredited” by anybody. This particular study is something else again, however.
I’m reasonably certain that some sort of systematic error crept into the data that Burnham, Roberts, ”et al.” collected in Iraq. I even have some idea how a systematic error might have been introduced by the unusually large cluster size, and I know how to test the data for that particular kind of bias, if I can get my hands on the details. I’ll try.
Oh, yeah — I did uncover one very interesting fact with all my reading. The Iraq Life Insurance Company has recently started issuing a one-year all risks life insurance policy. You can read about it here.
The mortality statistics from the Iraq Life Insurance Company would certainly be useful in furthering the discussion of the ”Lancet” study. The story in the ”NYTimes” said the company has already issued 2,700 of these new policies, and has yet to pay a single claim. I’m not sure exactly how long those policies had been in force when the story was written, but if the ”Lancet” figures are correct, then the current overall risk of death for a typical Iraqi is 1.98% — per year. So the company should get 54 claims on those 2,700 policies. Even without additional information about when the policies were issued the comparison of 0 claims to 54 expected seems striking.
A quick rule of thumb for one-year term life insurance is that the gross premium charged by the company is about twice the actual risk of loss on the policy — 50% of the premium typically goes to pay claims, and the other 50% represents the salesman’s commission, and an allowance for overhead expenses and profit. Since the actual premiums charged by the Iraq LIC range from 1.4% of the face amount (for low risk occupations like bankers and accountants) to 2.5% of the face amount (for high risk jobs like drivers, and translators for the Americans), it appears that the Iraq LIC estimates that the risk of death in Iraq ranges from 0.7% to 1.25%, after applying this admittedly crude rule of thumb.
Clearly the majority of the population lies in the lower risk categories, since almost 50% of them are under the age of 20, and probably not going to work at all. An overall current mortality level of 0.7% – 0.8% per annum squares pretty well with other sources of data. So the ”Lancet’s” estimate of 1.98% is way too high, in my opinion.
Anonymous said
David, thank you very much. Now, if you’d just start your own blog, you could present all the other interesting data you collected and shared with us at breakfast this past Saturday. 🙂
Folks, that breakfast would have made a great podcast, had someone recorded it. David researched historical Iraqi birth and death rates, population sex and age distributions, etc., etc., and found some fascinating facts (I wish I’d taken notes). From those, he made some interesting “back of the envelope” calculations that suggested the Lancet study’s numbers are between 5 and 10 times too high.
David’s calculations are roughly in line with another strongly anti-war, but conscientious source, Iraq Body Count (IBC). In my earlier comment, I linked to their press release on the Lancet study. For an in-depth look at the attempts by Ken Roberts (the Johns Hopkins researcher who’s primary author of both the studies published in Lancet) and his friends to discredit IBC and misrepresent the available data, see this detailed defense.
Anonymous said
It’s been a while since I attended Saturday breakfast. Is it still at lepeep on Colorado blvd?
Anonymous said
You’re right, Walter — it’s the LePeep at Colorado & Mexico. At 8 AM, you might spot the once and future blogger Jed Baer of FreedomSight at the round table in the back. The rest of us straggle in by 8:15. It would be good to see you again.
David Bryant said
”From those, he made some interesting “back of the envelope” calculations that suggested the Lancet study’s numbers are between 5 and 10 times too high.”
I don’t think I said 5 – 10 times. I’m not even sure I quantified it. I do think it’s too high by at least a factor of two. I’m certain I didn’t say 10 times. That would bring ”The Lancet’s” number in line with the one from IBC. And I’m sure that the IBC number is a lower bound, not a reasonable estimate.
I suppose almost everyone is sick and tired of this subject by now. There have been some interesting developments recently. A couple of physicists and an economist (all from England) have publicly criticized the ”Lancet” article. That argument made it into ”Science” magazine — you can read about the dispute over here. Oh — the piece I’m referring to is near the bottom of that page. You might want to search on the word “Lancet” to find it.
The Wall Street Journal ran an editorial on the subject. A blogger named Tim Lambert jumped all over that editorial, and eventually Steven Moore (who wrote the WSJ editorial) entered the discussion. You can find all of that on [http://scienceblogs.com/deltoid/2006/10/flypaper_for_innumerates_wsj_e.php#more/ Lambert’s blog]. Lambert is extremely partisan in his point of view, but he’s no dummy. Mr. Moore has been conducting polls in Iraq for several years now, so he understands the conditions on the ground fairly well. I don’t think these two reached any kind of understanding, but they do present the two sides of the argument pretty forcefully, and with minimal invective.
What else? There’s a dispassionate description of what the study says, and what it doesn’t pretend to say, on a blog that’s sponsored by George Mason University. If more people would think about statistics as carefully as Ms. Goldin does, the debate over this study would be a lot more edifying.
Finally, many of the people who attack the ”Lancet’s” study keep coming back to the question of death certificates: If there really are that many people dying in Iraq, why doesn’t the Iraqi Ministry of Health have a boatload of death certificates in some office building in Baghdad?
Most people don’t have a good grasp of how long it takes to compile decent statistics. For example, the U.S. CDC has only recently finished tabulating the death certificates for the year 2004. If data collection runs almost two years behind the event in the United States, why would anybody think it could possibly get done a whole hell of a lot faster in Iraq?
Anyway, leaving technical details like that to one side, there are other plausible reasons why many deaths in Iraq might go unreported. This news story in the ”Washington Post” provides a chilling insight into what some of those reasons might be.
David Bryant said
Just a quick note … I finally found a link to the version of this study that has been posted on [http://web.mit.edu/CIS/pdf/Human_Cost_of_War.pdf/ MIT’s web site]. This version of the study contains more detail than the version that was initially released by ”The Lancet”.
VRB said
Is there any reason not to believe that study. Civilians die in war and especially during guerrilla warfare. That count is almost insignificant compared to the number that have died in other skirmishes. Is the administration trying to say this isn’t a war?If medical technology was still the same as in Viet Nam, we would be probably looking at around 15,000 soldiers dead instead of 2800.
The study’s methodology appears resonable to me, if it so far off then something would have to be wrong with their sample. From what they said, their metodology had been tested with a similar sample size where they could verify the method from an actual count. If they were to skew the results to please anti-war avocates, what would be the point to go Iraq and put themselves in harms way. Why wouldn’t they just fudge it?
But, what do I know, I don’t live and breathe statistics.
David that link did not take me directly to the study. Try lowercase “CIS”