Count the dead in Iraq

How is he refusal to count casualties causing Iraqi deaths?

Knowing how many people are dying and the causes of those deaths is necessary to set priorities.
Given limited resoruces, is restoring public order more important than combatting infectious diseases?
Are islamic vigilantes and honor-killings a serious threat?

I have some real scientific issues with the Lancet survey which are independent of any political feelings I have about the war. (FWIW, I supported the war initially, and then turned against it once it was clear that the authors of the war had lied about the kind of reception the US forces would receive from a significant minority of the population.)
I am not trying to discredit the Lancet survey, which seems to have made the best of a difficult job. But it is possible that their headline figure of 100,000 war fatalities may have been skewed by statistical biases. I would like to know the answer to criticisms of two fundamental aspects of the Lancet study of Iraq's post-war civil demography: <ul>
<li>Underestimation of pre-war mortality</li>
<li>Overestimation of post-war mortality</li></ul>
Pre-war mortality: The CIA put Iraq's pre-war crude death rate at 6/1,000 (2001). THe WHO put it at 8/1,000 (2003). The World Bank estimated it a 8.8/1,000 (2000). Lancet estimated Iraq's pre-war death rate at 5/1,000. This is a 15%-80% underestimation from the typical international agency estimate. It compares badly with Lancets estimate of Iraq's post-war (2004) death rate of 12.3/1,000. Obviously, statistical conclusions about the the mortality effects of the war are going to greatly influenced by the choice of base year and agency source.
Post-war mortality: The Lancet study infers that most of the implied massive increase in mortality rates since the war is due directly to war-related violence or indirectly to the break down of human lifes-supporting services. How does this square with the relatively low figures of persons presenting as violently deceased in mortuaries?
Iraq Body Count puts total war-related civilian deaths at 16,000 - 18,000. This is closish to an order of magnitude less than the Lancets estimate of the extra war-related mortality of around 100,000 dead (exluding Fallujah).
Obviously the fact that mortuaries are what the Economist characterises as "passive surveillance" and will underreport. Field researchers are "active surveillance" and tend to get a more accurate sample.
But the difference in values is pretty staggering and a little hard to believe. A value of 100,000 killed implies a weekly total of about 1,000 week, week in week out for two solid years straight. That is about 150 killed per day. This would mean that the US and the insurgents are doing the equivalent of a My Lai massacre each and every week of the year. Perhaps if the Fallujah casualties are taken into account this is possible but it would be nice to have the reason for this apparent discepancy spelled out.
It seems more likely that the increased post-war death rate is due to a higher crime rate and social disorder. This is, of course, a result of the war and a consequence for which the US admin should be held responsible. The deliberate failure by US authorities to collect and calculate mortality stats is in iteself an accountability failure which is an indictment of the US government.

Jack Strocchi - first, thanks for a reasoned post. This is a welcome change from the usual nonsense that is written by the Lancet study detractors.

Regarding pre-invasion mortality: Mortality has been going down since the introduction of the food-for-oil program (this is evident in your numbers too, and is also something to remember when considering criticizm of the program). Therefore there is no contradiction between 2000 and 2001 estimates and the Lancet estimate. The WHO 2003 number covers mostly the post invasion period.

Regarding comparison to mortuaries and IBC: The mortuaries number is controlled by Iraqi/US authorities and as such is completely unreliable. The IBC should be compared to 50,000 violent deaths Lancet estimate rather to the 100,000 total. This gives a ratio of about 1:3 - which seems quite reasonable or even low. Do you really think that the press manages to track more than a third of the violent deaths in Iraq?

As for the 150 dead per day. First note that, again, only about half of these are violent deaths. Second, this number is not unusually high for the situation. The numbers in Vietnam were probably much higher. The purpose of the celebration of the My Lai massacre is to make you think that this was an isolated incident, but in reality it was just another run-of-the-mill event.

Thanks for that pro-bono mathematician. So to summarise: Iraq's pre-war mortality rate was in secular decline owing to the re-cuperative effects of the much maligned UN oil for food program. This makes the CIA's 6/1,000 (2001) look like a plausible pre-war baseline figure and it was trending downwards to the Lancets 5/1,000 (2002) figure.
The escalation in Iraq's post-war mortality figures is attributable, in roughly equal proportion, to both direct (martial violence) and indirect (civil negligence) war-related causes.
The military violence deaths are heavily underreported by politicised and passive surveilling government agencies. A figure of about 25,000 pa KIA is not implausible given comparable levels of violence in other sectarian civil wars where massive US firepower has been involved.
The level of civil negligence deaths has barely been reported at all, owing to their unspectacular nature. But they are occurring at a much higher rate owing to the post-war breakdown in human life-support services as evinced by the increase in child malnutrition. This is caused by the incompetence, rather than malevolence, of the US admin.
Things may turn around, especially if the Suuni minority accept some kind of consensual national civil government. And one must not forget that Hussein's violence and incompetence were responsible for occasional large scale spasms of mortality (due to war, repression & national isolation). Still, it is hard to resist the conclusion that the Iraq war has been a humanitarian-moral, as well as utilitarian-political, disaster.

I also have some issues with the Lancet study; it's pretty clear that the sample size is too small to estimate the pre-war violent deaths. One issue I worry about is people's recollection - certainly they will recall pretty precisely if their family member died pre-war or post-war, but will they remember and report accurately deaths around the start of the survey period, which was three years prior to the survey date.

Also it is possible (but unlikely) that the survey is simply an outlier. I'd really like to see th results of an equivalent survey conducted again.

Patrick,

I am not an epidemiologist, but based on my layperson's knowledge:

1) I don't believe the sample size was too small at all. In fact, as I recall, 32 clusters is considered quiet enough for these kinds of studies, and the authors of the study included 33 clusters, so the sample size was actually somewhat larger than necessary.

2) Did you read the paper that was published in The Lancet? If not, I suggest you do, since the authors did discuss the issue of recall accuracy.

3) I don't understand your concern about accuracy at the start of the survey period, which was not 3 years prior to the survey, but around 1 1/2 years. What would be inaccurate about their recall, and why?

4) I am not sure why you would think the survey may be "simply an outlier". I can see nothing about the methodology or any other aspect of it that would lead one to that notion.

5) The authors of the study would like to see more studies done as well,and so would I. I'm not sure how likely that is as long as the Bush administration's occupation controls information. This is certainly something they want to keep a lid on if at all possible.

Sorry - I thought I had made the correction, but for some reason it didn't stick. The start of the survey period was 2 1/2, not 1 1/2 years before the survey.

1+2) As it happens I read the paper, but now a few weeks ago. The statement I made was that the sample size was too small to estimate the pre-war violent deaths; as I recall there was one pre-war violent death recorded. I can't recall the article's statement about this, but I'd be sure that this is sufficient to put an upper bound on the number of violent deaths pre-war, but not get it into a range of say a thousand.

This is important since one of the repeated statements about the war is to the effect that Saddam was a horrid murderer, killing thousands (or hundreds of thousands, depending on hyperbole used) a year. It would be good to have an estimate of this number backed by some kind of data.

On 3), the survey requires the participants to accurately recall the time of death of people in their household. There could be some bias around the start of the survey period - I can't for instance recall precisely when my grandmother who died about 4 years ago did die, except that it was in the winter sometime. There's then a possibility of miscounting at the start of the survey periods. I did misquote the length though.

4) The statement that the survey is "simply an outlier" is badly worded sorry. It is possible (but unlikely) that the survey is simply wrong; there is (for instance) a 5% chance that the true number of deaths is outside their confidence interval. Any statistical survey suffers a flaw of this nature. The estimates produced by the survey are almost orders of magnitude greater than those from other sources. There is a small probability that the cluster sampling happened to come across areas of high concentrations by chance.

I suppose I could put it this way; prior to reading this survey, I would have said that it was practically impossible that the true number of deaths by violence was 98,000, because most estimates came in at under 20,000. If I had only read the survey I would have estimated the probability of 100,000 deaths (or thereabouts) as being pretty high. I don't know how to reconcile these and I find the differential counting explanations (of reported as against actual deaths) to be a bit facile. Maybe this was just the 5% "bum" survey?

5) I'll only note that the first study was done despite the machinations of the Bush administration, and a second study would seem to be a pretty good way of getting one's name into Lancet, so I'd be surprised if someone hasn't started of it already ...

Jack
Not only does the IBC count omit deaths from disease etc it also excludes deaths of combatants (on all sides). It is, as the site itself says, a record of civilian deaths by violence in the period following the invasion.

This is important since one of the repeated statements about the war is to the effect that Saddam was a horrid murderer, killing thousands (or hundreds of thousands, depending on hyperbole used) a year.
Saddam averaged 70K/year between 1988 and 2003, but this is heavily loaded on the period 1988-1992 (the later stages of the Iran War and the massacre of Shias immediately after GW1). For the period under survey, it was more like 2,000/year.
Patrick: on your 5), I wouldn't hold your breath. Doing another survey like this would be a good way of getting in the Lancet, but it would also be a good way of ending up dead. The original survey team were quite clearly in favour of their lives, and AFAICT the security situation in Iraq has got materially worse since the fieldwork was carried out.

It strikes me that one of the things the Lancet article needs badly is a table stating something like there is an x% chance that the actual number of excess deaths exceeded 10,000, 20,000, 50,000 and 100,000 (well, we pretty much know the first (~2.5%) and the last (~50%), but the other numbers would be useful in trying to put the study in context.

You can make one in Excel with the formula =1-(NORMDIST(N,98000,D12,TRUE)) where N is the number concerned.

(Note to statistics pedants; of course that isn't going to give the correct sampling distribution, but nothing else is either and I say that this normality assumption isn't too bad).

Kudos to this blog for keeping the topic alive - thanks, Tim.

For the period under survey, [Saddam's murder rate] was more like 2,000/year.[dsquared]

I haven't seen this claim before. What's it based on?

The US State Department, amnesty International and Human Rights Watch all prepared annual reprots on Iraq in the period prior to the 2003 invasion.

Here's an excerpt from the 2002 Human rights Watch report:

Government opponents and relatives of political detainees continued to report numerous executions of political suspects and those convicted of ordinary criminal offenses, as well as former army personnel suspected of disloyalty to the authorities. Scores of civilians detained in Abu Ghraib prison were apparently executed in early March, among them five Shi'a Muslims from al-Najaf province who were arrested in December 2001 after reportedly failing to cooperate with the authorities in the capture of army deserters. Their bodies were returned to their families later in March. Ten other suspected government opponents were also executed in Abu Ghraib in June, among them Jabbar Sadeq 'Ali and 'Abd al-Salam Hadi Jawad, both from Basra. The bodies of these victims were apparently not returned to their families, and were buried by the authorities in a Baghdad cemetery. A number of armed forces personnel were also reportedly executed by firing squad in Baghdad, Mosul, and other cities in March, among them 'Abd al-Haq Isma'il and 'Abd al-Hussain Jassim.http://www.hrw.org/wr2k3/mideast4.html
Precise numbers are unavailable for obvious reasons but estimates of killings attributable to Saddam's regime in the period 1998-2002 are typically in the high hundreds or low thousands per year.

Can anyone tell me what period was covered by the Lancet study?
I understand the study was conducted in September 2004 but since the study is no longer freely available on the Lancet site I can't see what period respondents were asked about.
Iraqi Body Count reports roughly 2100 deaths for the period since 1 October 2005. That's around 13% of the total (minimum) bodycount they estimate.
As discussed earlier in this thread the IBC number isn't directly comparable with the Lancet figure. However it seems clear that there have signifcant additional casualties in Iraq since the end of the Lancet study.

You can find a copy of the Lancet study here.

The HRW 2003 report gives the impression that they can only verify dozens of executions not hundreds or thousands.

Naturally, as a typical "let's blindly take swipes at the Bush administration" leftie, you completely discount the value of a future democratic state, one that will save countless lives, not to mention livelihoods. Even if the Lancet figures were close to accurate, and they've been proven false over and over, they would still justify the Iraq liberation. Unless, of course, you're a compassionless leftie crank who cares nothing for the liberty of others. You don't deserve the freedom you have, nor do you recognize its significance. You're blinded by irrational hatred of one man (or one party... your tired song and dance varies slightly from week to week).

Be a man.

Nash, you forgot to close your [/sarcasm] tag when you hit p0st. Not to worry, I closed it.D

Dsquared, where does the average 70K a year for 1988-2003 come from? That's a million--are you including sanctions deaths? Saddam does share some blame for those, along with the US and UK.
I've seen the figure of 300,000 killed by Saddam cited a bunch of times, and if one included Iranian and Iraqi dead in their war (but that was over in the period you cite) you might get over a million.

Donald,

You are passing the buck. According to Denis Halliday and Hans von Sponeck, two senior UN officials who were in charge of the UN oil-for-aid programme, the blame for the sanctions deaths lies ENTIRELY with the U.S. and U.K. goverments. Halliday recently stated that 'they' (meaning the US/UK governments) will be 'slaughtered in history for their complicity in genocide'. Certainly, Saddam used the sanctions to strengthen his dictatorship, but the food and medicine that was illegally blockaded had been distributed earlier very efficiently to the Iraqi civilian population, according to both Halliday and von Sponeck. Madeline Albrights famous words, that "the costs [in dead Iraqi children] were worth it" baldly reveals western complicity in this crime. Haliday claimed that up to 6,000 children were dying a month in Iraq at the height of the sanctions. And if you want to tally up victims for comparison what about Gulf War I? Several estimates of Iraq dead range as high as 200,000, and let's nt forget the highway of death and the carnage that occurred there. Finally, its a bit rich for people to whell out the Iranian death toll when this slaughter was fully supported by the Reagan and Thatcher governments, who pushed Saddam to keep up his war with Iran. Again, so many convenient facts go down the memory hole....

dsquared 15/3/2005 04:47:34 wrote
(Note to statistics pedants; of course that isn't going to give the correct sampling distribution, but nothing else is either and I say that this normality assumption isn't too bad

Well, I am a pedant, but statistics is not my obsession. My worry was that in a study of the Lancet type, the distribution would not be normal with wider wings so that using the normal distribution would overestimate the central tendency. As a general rule does anyone have a feel for this.

The Lancet study has the most likely number of deaths at between 8,000 and 198,000.

Pick a number.

Such a large interval (better than 20:1) does not give much confidence in the study.

Jeff,

I take it you are not bothered by the UN & Co. stealing food from the hungry.

The Kurds got the same $$ per capita as Saddam did and they had no problem with excess deaths. Why?

M.Simon, This is a smokescreen and the Americans know it. The sanctions resembled a medieval siege and were aimed at harming the civilian population of Iraq in primarily Sunni areas. If you bothered to read some of my earlier posts I quote an American general who admitted so after the coalition bombed the country's civilian infrastructure almost flat in Gulf War I (but not so much in Kurdish regions; the central and southern areas were by far the hardest hit by coalition bombs). But if you want to believe in the lies that our media drip feed us over the arguments of Halliday and von Sponeck who were there and in charge of overseeing the distribution of aid, then that is your problem. Furthermore, why was there no media coverage of Turkish aerial attacks on Kurdish villages in Iraq on at least two occasions, during the time that the US/UK maintained a no-fly zone over Iraq? According to some US pilots, they were told to stay grounded while the Turkish air force got on with the killing. Why has this been universally ignored by our media? Obviously, its because they have a role as protectors of the established order. When that it even threatened only a little, they rally around it with distortions and lies. Lastly, you have no dea what 'confidence intervals mean'. You make an argument that has been so systematically demolished that defenders of western aggression and illegality are now forced to attack the sampling methods. The most likely death total was, according to the study, just under 100,000 excess deaths. As the death total deviates from that figure (higher or lower) it becomes increasingly unlikely as an accurate figure. Moreover, if the slaughter carried out in places like Falluja and Najaf was included, the total would probably be far higher than 100,000 dead.

No Iraqi names (or other arab nationals or any non-westerners) are listed on the signatories, yet there's a Canadian listed there. I suppose it's because any Iraqis who criticise the coalition will be hunted down by Karl Rove's death squad.

The letter hardly sets out a convincing case that it has to be the US/UK that has to do the counting.

I suspect that some people (mainly in the arab world) claim the US/UK deliberately targetted civilians. If the US/UK were to count the civilian deaths, do you think this would increase or decrease the number claiming that?