My Tiny Contribution to the Debate on the Lancet Report I Have Not Read
One of the objections you see floating out there is that the jump in the median excess deaths estimate from 100,000 in the first report to 655,000 in the second report is just too big to believe. But, hokay, each report had a confidence interval. The confidence interval on the first report was something like 8,000-200,000 while the confidence interval on the second is 400,000-975,000 or thereabouts.
Critics of Lancet I saw the confidence interval and said, “A ha! So there could have been as few as 8,000 excess deaths!” They did not stress that there could have been as many as 200,000, even though, statistically, the 200,000 number was as likely as the 10,000.
Now, if the “real” excess deaths during the period of the first report was close to the upper bound, and the “real” excess deaths during the period of the second study is close to the lower bound, then your jump is from 200,000 to 400,000, not from 100,000 to 655,000 (or 10,000 to 975,000).
That’s not nearly so dumbfounding, is it? The excess deaths double during a period in which even Ralph Peters might acknowledge that the violence has gotten worse.
Comment by Ron Eggleton —
October 15, 2006 @ 10:21 am
Les Roberts, one of the original authors of that Lancetreport (of 2004), is known to have said this past February that the number of Iraqi casualties could be over 300,000, so the jump should not come as that much of a surprise.
This is from Dahr Jamail, someone who has been a witness to the destruction in Iraq outside of the protection of the US military:
http://www.dahrjamailiraq.com/hard_news/archives/newscommentary/000476.php
Comment by Tim Lambert —
October 15, 2006 @ 11:30 am
Actually they say that the second study gives about 110,000 excess deaths for the period matching the first study. The increase is due to a huge increase in violence since the first study was carried out.
Comment by Neel Krishnaswami —
October 15, 2006 @ 11:41 am
I’ve skimmed it, and I don’t believe that they the methodology they described is the one that they actually ended up happening on the ground. The idea that the poll takerers would be able to travel safely into randomly selected streets seems implausible, unless they got permission from the local bosses ahead of time — and that means that we can’t tell if the people who were polled were really randomly selected or not.
However, the important fact isn’t the particular numbers — it’s that Iraq is such a violent and lawless place that this is probably the best possible study that could be done. Iraq is too dangerous a place to even merely count the dead, which ought to embarass any remaining apologists for the invasion.
Comment by Gnorgathon —
October 15, 2006 @ 1:44 pm
That warrants being repeated.
Comment by matthew hogan —
October 15, 2006 @ 2:45 pm
The interpretations of the study — with the additional big proviso that counting dead in this environment and getting reliable informants (people who will honestly say who actually killed the relative) and surveyors has problems — almost wilfully confound the results.
Excess deaths mean any death overage that might be attributable to the circumstances set up since the invasion. Violent deaths can include banditry, feuds, the usual personal motives, as well as sectarian, and coalition war violence. And who was killed as a fighter or civilian.
The numbers again seem way high. I still feel Iraq Body Count is most reliable and one can add maybe 100% more to account for information lost in the less travelled areas.
Until the civil war really expanded after the big Shiite mosque attack it appeared there were about 10 Iraqis (insragent and civilian) dead for each GI; current napkin-memory calculated trends seem to suggest about 30-100 Iraqis dead for each GI dead, as the civil war expands with Iraqis killing Iraqis, and GIs getting out of the way. This could mean war casualties are entering the 100,000s. But dont forget that that means the entire conflict would have had 1000 dead per day on average which doesn’t match the facts over most of the past days and not the past years.
The first Lancet study contained an obvious absurdity. It used select areas and in its finding it reported that 2/3 of the deaths there occurred in Fallujah.
If that figure was indeed about 100,000 then 60,000 died in Fallujah up til them, which is something like 25% of the entire city!
Dahr Jamail’s heart may be in the right place but that doesnt mean his critical mind is assessing everything.
Lots of unnecessary deaths and murders and cruel war violence do not require facts out of accord with reported reality.
Comment by Alex —
October 15, 2006 @ 3:14 pm
I’d say it bears yet another repetition.
Comment by matthew hogan —
October 15, 2006 @ 3:25 pm
And another.
Comment by Alex —
October 15, 2006 @ 5:32 pm
So, to conduct this study, did they walk through Iraqi cities calling “Bring out your dead! Bring out your dead!”?
And what do they do if somebody brings out a relative who isn’t dead yet, and insists that he feels fine?
Trackback by Inactivist —
October 15, 2006 @ 5:38 pm
Bring out yer dead!…
Like Jim Henley, I haven’t read the recent study of Iraqi casualties. I know almost nothing about the methods (well, I know statistical theory, but I know very little about the practice in contexts like this), so I won’t bother commenting on the val…
Comment by Barry —
October 15, 2006 @ 6:29 pm
Matthew Hogan: “The numbers again seem way high.”
Matthew, if you’ve been following the controversy at Deltoid (scienceblogs.com) and/or Crooked Timber, you’ve seen a lot of people come up with that argument from incredulity. Frankly, it’s justifiably a suspicious argument, since neither you (I hope) nor I (thank God!) have any experience at all with civil wars. This is arguing ‘that seems too high’ from not even distantly related experience. What we do have is some knowledge of recent civil conflicts. From the Lancet article (references in the article): “In Iraq, as with other conflicts, civilians bear the consequences of warfare. In the Vietnam war, 3 million civilians died; in the Democratic Republic of the Congo, conflict has been responsible for 3·8 million deaths; and an estimated 200 000 of a total population of 800 000 died in conflict in East Timor. Recent estimates are that 200,000 people have died in Darfur over the past
31 months.”
Matthew: “I still feel Iraq Body Count is most reliable and one can add maybe 100% more to account for information lost in the less travelled areas.”
The Iraq Body Count, as is well known to anybody following the Lancet arguments (first or second survey) has as its criteria “Casualty figures are derived from a comprehensive survey of online media reports from recognized sources. Where these sources report differing figures, the range (a minimum and a maximum) are given.”
(http://www.iraqbodycount.org/background.php). This means that deaths are required to be reported by media sources, online ones at that (I’ve heard, but don’t know, that they also require verification from a second source).
This means that the Iraq Body Count is a *hard lower bound* on the *violent* death figures. It was mentioned, with references, in the Lancet report that, as violence escalates in previous civil wars, the percentage of deaths being reported would go from 50%
down to 20% or lower (with the notable exception of Bosnia). This implies that the lower bound on violent deaths is from 2.5x – 5x the Iraq Body Count figure, which would be 125,000 – 250,000. And that’s just violent deaths. And it’s assuming that covering the Iraq War is no difficult from covering other wars.
Comment by Nell —
October 15, 2006 @ 7:40 pm
Also, Matthew, you’re simply mistaken about the first Lancet study and Fallujah. That study explicitly excluded Fallujah from sampling because the amount of fighting there had made it an outlier.
Comment by matthew hogan —
October 15, 2006 @ 9:09 pm
Nell — I’ll recheck, but I remember it being odd the way they handled fallujah.
But isnt that sort of like excluding the Great Depression from economic growth statistics because it’s an outlier? Anyway, need to check.
Barry — I have crunched war numbers a bit in my time, generally to conclude that wars are MORE vicious and wilfully cruel than the conventional wisdom. But numbers do matter, and use of firepower, and wilfullness. Wars dont just kill civilians willynilly, contrary to widespread myths. It would be nice if the study publicized itself with clear breakdowns of what excess deaths occurred where.
It’s true that media-reported deaths are used by Iraq body count, and that’s good methodology because any numbers can be floated by anyone. The fighting has generally been in the populated better-reported areas, and few specific regions, where coalition forces intersect with SUnnis or Sunnis and Shiites mix or Arabs and Kurds. They may be off but probably not by a huge amount.
Comment by matthew hogan —
October 15, 2006 @ 10:31 pm
First Lancet study:
“Two-thirds of all violent deaths were reported in one cluster in the city of Falluja. If we exclude the Falluja data, the risk of death is 1.5-fold (1.1-2.3) higher after the invasion. We estimate that 98000 more deaths than expected (8000-194000) happened after the invasion outside of Falluja and far more if the outlier Falluja cluster is included.. . . Violent deaths were widespread, reported in 15 of 33 clusters, and were mainly attributed to coalition forces. Most individuals reportedly killed by coalition forces were women and children. . . . INTERPRETATION: . . . Violence accounted for most of the excess deaths and air strikes from coalition forces accounted for most violent deaths.”
Now we have some idea of actual population (not sample) coalition-induced deaths in Fallujah up to September 2004 which is PRIOR to the heavier (maybe 5-8000?) death toll in the November 2004 reconquest; (this must have included the April 2004 post-mutilation aborted attack). Fallujah also saw some violence including two US troops shootings into crowds, which was startling in the number of a dozen.
If Fallujah was 2/3 of violent deaths to that point, even if we use the figures from the more ruthless November crackdown, we get a total for Iraq of around 10,000 deaths at that point, if Fallujah was two-thirds. That’s probably too SMALL back then.
The point is when one gets a figure like that, so out of joint with reality, the methodology should be questioned. As a matter of funny red flag numbers.
Lancet results are saying that to date, the death toll is nearly as high as all combatants in the five year US civil war.
Now allowing for infrastructure failure and high stress and increases in disease death and all that, maybe there is a large “excesss death” thing in their range.
But also quoting such high figures inures people to believe that a “mere” 40,000 is NOT a high number. It is. Civilians die because a war is waged directly on them, and a stupid war with all the risks of popular guerilla resistance and the counterinsurgency violence and ultimately ethnic terrorist strife.
People get too inured to the idea that “oh civilians die in war” and a few thousand is just natural incidental risk. Therefore we need to establish 100,000 or more.
(Personally I doubt the 4 million figure from Vietnam, if that is meant as the number of those died by hands of shooting and bombing warfare. Perhaps, of causes related to general displacement and loss, but one still has to rely on a 400 dead per day figure which would mean a My Lai massacre level of death every day for the whole conflict.)
Comment by BruceR —
October 16, 2006 @ 2:32 am
Matthew, you’ve completely missed the point of the Fallujah exclusion in the first Lancet study.
The point was one of their clusters happened to be in Fallujah: it was also their only cluster in Anbar province for that matter. It was obviously going to be an outlier, so it was excluded because it would have completely thrown out the numbers on a national basis if it had stayed in.
There is no way you can make the kinds of extrapolations (about how two-thirds of all Iraqi fatalities were in Fallujah) you’re making, from one cluster of families in one neighborhood from one cluster, and the Lancet rightly recognized that at the time.
This newer study uses a much larger sample size, eliminating some of the problems with error margins the first study had. It also uses a new method of defining excess deaths, which the first one did not. Tim Lambert is right in #2, saying that the numbers for the same time period from the second study are actually remarkably close to the number from the original study, so I’m afraid Jim’s point of view, above, would not be one shared or supported by its authors.
Comment by BruceR —
October 16, 2006 @ 2:35 am
Also, Matthew, 1,000 dead per day is 1.2 million over the course of this latest study’s period. That’s 50% again above their CI95 higher bound, so I don’t see your basis for claiming that’s the upshot of their result.
Comment by Barry —
October 16, 2006 @ 8:08 am
And Matthew, as the Lancet article noted, civil wars can kill in the multiple 100k level. This makes 600K from Iraq ‘credible’, whether or not true, not ‘incredible’.
Comment by Wild Pegasus —
October 16, 2006 @ 9:39 am
As a sidenote, the US War Between the States lasted almost exactly four years: 04/61 – 04/65. During that time, about 600,000 soldiers died, as did about 300,000 civilians. So the Iraqi number is certainly possible: the US invaded 4.5 years ago.
- Josh
Comment by doubled —
October 16, 2006 @ 2:42 pm
wild pegasus says: “As a sidenote, the US War Between the States lasted almost exactly four years: 04/61 – 04/65. During that time, about 600,000 soldiers died, as did about 300,000 civilians. So the Iraqi number is certainly possible: the US invaded 4.5 years ago.”
Where are the major battles in Iraq that could have caused such numbers to be comparable? Where is a battle such as Shiloh or Gettysburg where tens of thousands died in a single battle? The press hasn’t reported any battle with such casulty numbers that I’ve seen. Yet , some are still inclined to believe these statistical gymnastics?
Comment by Wild Pegasus —
October 16, 2006 @ 6:36 pm
You don’t need major pitched battles when the common people themselves are engaged in the violence. To wit, the Battle of Gettysburg killed fewer than 8000 people.
- Josh
Comment by Barry —
October 16, 2006 @ 6:38 pm
Where was the Shiloh in Rwanda? Bosnia? East Timor? In Rwanda, they broke the myth of Africans not having the work ethic for modern serious mass murder, yet I don’t recall any 10K death battles. There might have been, but I don’t recall any headlines.
As for Iraq:
http://www.macleans.ca/topstories/world/article.jsp?content=20061016_134735_134735
(1,000 per week, at the Baghdad morgue alone)
http://www.telegraph.co.uk/news/main.jhtml;jsessionid=HLKNZLDDOQO1TQFIQMGSFFWAVCBQWIV0?xml=/news/2006/03/31/wirq131.xml
(30K refugees, one month, Baghdad area)
The big difference between Iraq and the US, as far as civil wars go, is that there’s a lot more intermixture. This means that, in several provinces, there’s not a separation, with fighting along the border area, but a dispersed and continual violence.
“Yet , some are still inclined to believe these statistical gymnastics?”
When you have the training to begin to understand the statistics involved, please come back. Heck, when you’ve read the article and comprehend it, please come back.
Comment by simon —
October 17, 2006 @ 12:09 am
Iraq Body Count has a critique of the new Lancet study.
Comment by Barry —
October 17, 2006 @ 9:39 am
http://leninology.blogspot.com/2006/10/counting-dead-ibc-attempt-to-undermine.html
Is a good discussion; it seems that IBC has forgotten that their data comes from media reports (i.e., it’s hard lower bound on deaths). Or they haven’t forgotten, and are pissy to the point of lying, about people stepping on ‘their’ turf.
Comment by Barry —
October 17, 2006 @ 10:22 am
From Wikipedia (http://en.wikipedia.org/wiki/Battle_of_Gettysburg#Background_and_movement_to_battle): “Lee gave strict orders to his army to minimize any negative impacts on the civilian population. Food, horses, and other supplies were generally not seized outright, although quartermasters reimbursing northern farmers and merchants using Confederate money were not well received. Various towns, most notably York, Pennsylvania, were required to pay indemnities in lieu of supplies, under threat of destruction. The most controversial of the Confederate actions during the invasion was the seizure of some forty northern African Americans, a few of whom were escaped slaves, but most of them freemen, sending them south into slavery under guard.[9]”
Now, imagine the results would have been like if Lee didn’t have a large infantry force, just a much smaller cavalry force. However, his goals wouldn’t have been to fight other armies, but to kill as many civilians as possible. And, in many cases, it wouldn’t have been people going hundreds of miles to kill, but to the next neighborhood or village.