Tim Lambert (deltoidblog AT gmail.com) is a computer scientist at the University of New South Wales.
A new survey puts the Iraqi death toll at over one one million:
These findings come from a poll released today by O.R.B., the British polling agency that have been tracking public opinion in Iraq since 2005. In conjunction with their Iraqi fieldwork agency a representative sample of 1,461 adults aged 18+ answered the following question:-
Q How many members of your household, if any, have died as a result of the conflict in Iraq since 2003 (ie as a result of violence rather than a natural death such as old age)? Please note that I mean those who were actually living under your roof.
None 78% One 16% Two 5% Three 1% Four or more 0.2%
Given that from the 2005 census there are a total of 4,050,597 households this data suggests a total of 1,220,580 deaths since the invasion in 2003. ...
As well as a murder rate that now exceeds the Rwanda genocide from 1994 (800,000 murdered), not only have more than one million been injured but our poll calculates that of the millions of Iraqis that have fled their neighbourhoods, 52% have moved within Iraq but 48% have crossed its borders, with Syria taking the brunt of refugees.
My back of the envelope calculation puts the 95% confidence interval at 1.1-1.3 million. This seems consistent with the second Lancet study giving 600,000 violent deaths when you take into account the amount of time that has passed since then. The surveyors in this study did not verify the deaths by asking to see death certificates, so this could bias the results upwards, but the experience in the two Lancet studies suggests that the great majority of people who report a death can verify with a death certificate. The number of households in Iraq has dropped by maybe 5-10% since 2005 because of all the people that have fled the country, so 1.2 million may be a little high, but it's likely that the number is now one million dead and another million injured as a result of the war.
The LA Times has picked up the story here. Comments from lenin and Stephen Soldz.
Politics
Comments
Interesting stuff. I have a high opinion of the ORB folks.
1) I agree that, given this sample size, the confidence interval would be quite narrow, although I can't tell from the write-up how much clustering was done. I assume that Tim's back-of-the-envelope calculation assumed (non-cluster) random sampling.
2) Would the math work out to a death rate of 40-45k per month since L2? 600,000 more deaths in the last 14 or so months. That would be 1,500 war-related deaths every day since July 2006.
3) What would the daily death rate be in Baghdad over the last year, assuming that both L2 and ORB are correct?
Even though I am a Lancet-skeptic, I try to be open to new information as it comes in. There is no doubt that this result is consistent with the Lancet results. Adjust your posteriors accordingly!
Posted by: David Kane | September 14, 2007 11:34 AM
Glad someone did this. It's been infuriating, the number of polls done in Iraq with almost no attention given to the casualty rate (with the partial exception of that poll conducted earlier this year which asked about "persons harmed" in a given household.) You'd almost think people weren't interesting in knowing the answer.
Posted by: Donald Johnson | September 14, 2007 11:35 AM
Tim What relationship does this firm have to the BBC? Do you know?
They're implying approx 840 extra deaths per day in Iraq as a result of the violence! Isn't a little common sense required at times?
Posted by: jc | September 14, 2007 11:42 AM
LA Times is reporting: According to the ORB poll, a survey of 1,461 adults suggested that the total number slain during more than four years of war was more than 1.2 million.
I don't think there has ever been a day reporting 840 slayings.
Posted by: Jc | September 14, 2007 11:50 AM
Donald,
I think it was not so much that people weren't interested in knowing the answer - plenty of people surely were - but that the thugs who controlled the Ministry of Health didn't want people calling their blatantly bogus figures into question. If that was the reason why so little data emerged, this study clearly indicates that the political situation has changed. Evidently it now suits the powers-that-be to allow the truth to come out. Why the change? I don't know.
Posted by: Kevin Donoghue | September 14, 2007 11:52 AM
Kevin, I was thinking more of the various polls that have been conducted in Iraq by Western groups--the ones we're always reading about which tell us how they feel about their government, the occupation forces, whether it's acceptable to attack occupation forces, etc... If the news organizations funding these polls wanted to estimate the death toll, they could have done this earlier.
Posted by: Donald Johnson | September 14, 2007 11:56 AM
Of course, there may also be a small error introduced by the fact that you can't poll a household that's been wiped out entirely...
Posted by: Dunc | September 14, 2007 11:56 AM
"What relationship does this firm have to the BBC? Do you know?"
Easy. Gordon Heald has worked for Margaret Thatcher, President Reagan and Boris Yeltsin. The BBC has interviewed Margaret Thatcher, President Reagan and Boris Yeltsin. Stands to reason, dunnit?
Snark aside, your point is what, exactly?
Posted by: Kevin Donoghue | September 14, 2007 12:03 PM
I don't think there has ever been a day reporting 840 slayings.
Why would there ever be a day when all deaths got reported, whether by the decimated media, the death-squad loving Ministry of Health, or by anyone else?
Posted by: Ron F | September 14, 2007 12:10 PM
The ORB Xcel spread sheet that shows the weighted figures suggests 1.2 million dead and 1.1 million injured. Has there ever been a war in history--seperate from organized genocides with concentration camps--where there have been more deaths than injuries? Typically in counterinsurgencies and civil wars there are at least 2-3 injured for every death.
The tables also reveal 264,000 car bomb deaths, most of which were inexplicably missed by the media (even though the whole point of car bombs is to call attention to them and even remote bombings of villages--like the recent attack against the Yizidis in northwestern Iraq--tend to be reported for this reason). ORB also estimates 347,000 injuries from car bombs. Does this low ratio of wounded to deaths make sense? Car bombs injure a lot more people relative to those they kill (from blast, shrapnel, etc.).
Ditto with air strikes: 116,000 deaths (mostly unnoticed despite insurgent/militia/anti-occupation incentives to publicize them) and 132,000 injuries from air strikes. Given the nature of aerial bombardment, does this under-reporting or ratio make sense?
A few other strange things from the detailed tables on ORB's website:
It looks like they oversampled Baghdad relative to its share of the population. This could be important in pushing numbers up.
Anbar is not included (as far as I can tell from the table listing provinces). Strange.
They undersampled Kurds and Shia Arabs relative to their percentage of the population, which would push numbers up given that violence in Iraq spiked last year when Shia militas began to cleanse Sunni neighborhoods in Baghdad, Diyala, etc.
They also interviewed a disproportionate share of non-Muslims (only 45% of respondents were "Muslim" overall and 28% percent in Baghdad -- the reset where orthodox, catholic, protestant, etc.). This is weird, but could conceivably push numbers up since religious minorities have been targets of some pretty terrible violence.
The study suggests a lot of movement/displacement. External refugee flows probably mean that ORBs baseline population estimate for Iraq is too high, and internal displacement has probably increased the average size of remaining households as people flee to live with relatives. Together, this would alter the total number of estimated households in Iraq downward and increase the average household size -- suggesting the ORB estimate is probably skewed high.
For all these reasons, it would be good to know more about the methodology here.
Posted by: CK | September 14, 2007 12:14 PM
"If the news organizations funding these polls wanted to estimate the death toll, they could have done this earlier."
Donald, I really think you are being a bit harsh here. It's very clear that some very ruthless people, who control (after a fashion) an assortment of militias in various parts of Iraq, really did not like the idea of independent researchers asking questions about mortality. Asking about opinions is not the same at all. Why resistance to letting the truth come out has softened I really don't know.
Maybe it signals a new phase in the war. John Quiggin wrote an interesting post about this - there is often a stage in a conflict when the story changes from "Things are not nearly so bad as the media says" to "We are killing the enemy in huge numbers". I'm just speculating, but it seems to me a new policy is in place, otherwise this study could not have been done as openly as it evidently was.
Posted by: Kevin Donoghue | September 14, 2007 12:19 PM
These data can't be used to calculate a mortality rate. David Kane is thinking "whew! Thank goodness for that!"
I'm not sure where they get the 4.05 million HH number.
The ratio of reported deaths to injuries > 1.0
Posted by: Robert | September 14, 2007 12:19 PM
I would argue that the survey may undercount deaths a little. There are many Iraqis that have no household, no roof, no phone anymore. They probably haven't asked the Iraqi refugees in Syria if they had a death in the family.
Posted by: _Arthur | September 14, 2007 12:24 PM
CK:
Let me go out on a limb here and guess you don't speak Arabic.
Posted by: grh | September 14, 2007 12:29 PM
grh: You are correct, although I read Arabic press in translation on Iraq almost every day. The 116,000 death figure for airstrikes works out to an average of 72 deaths every day for 1,600 straight days. Please forward stories from the Arab press suggesting this level of fatalities from any given day (since the end of declared major combat on May 1, 2003), let alone day after day after day after day . . . I have not seen such stories, but I'm open to being proven wrong.
Posted by: CK | September 14, 2007 12:42 PM
CK -
264,000 car bomb deaths, most of which were inexplicably missed by the media
Given that there's a brutal war underway, with the highest casualty rate among media workers in any conflict in the modern era, I think the onus is on you to explain why you think the media would capture any but a tiny portion of car bomb (or any other) fatalities.
CK -
even though the whole point of car bombs is to call attention to them
Why would Baghdad, Washington, or the largely supine media want to draw attention to evidence - car bombs, lot's of them - of the complete failure to bring security (which entails fewer car bombs)?
Posted by: Ron F | September 14, 2007 12:50 PM
"Why would Baghdad, Washington, or the largely supine media"
Definition of supine- "spineless"
Referring to:
NY Times Wash. Post LA Times
Yep the big american dailies sure are spineless when it comes to the bush administration. The ABC and NBC too.
They're all in the pockets of Hallibuton is my guess.
Damn Cheney. He's keeping this story out of the press again.
Posted by: Jc | September 14, 2007 1:00 PM
Ron F: Actually, my point was that the insurgents who conduct car bomb attacks do them to inflict mass casualties AND call attention to them -- they don't want them to go unnoticed and they have ready access to Iraqi and Arab regional media. The highest recorded death toll from one days bombings, 500, was the recent attack on the Yizidis (reported by Arab AND Western press despite being in a remote and isolated portion of northwestern Iraq). The highest level reported incidents before this are in the range of 150 -- and there haven't been that many. Yet, if ORB is right, there have been, on average, 165 car bomb fatalities every day for 1,600 straight days. If you are correct that the Western media (including media outlets in Europe that HATE this war) are systematically ignoring these deaths or are incapable of operating in Iraq (even with their Iraqi national staffs), please forward evidence from non-Western sources (the hundreds of insurgent websites, Arab media, Iraqi newspapers, etc.) that support this level of carnage.
Posted by: CK | September 14, 2007 1:02 PM
A few somewaht rambling thoughts.
I think that, in a sense, confidence intervals are beside the point, though they will certainly be affected by clustering. As I remember, the design effect in both L1 and L2 was less than 2. So, not that bad.
But, I assume that there is a major amount of nonsampling error in these surveys. In this case, they were unable to even conduct the survey in three provinces. Given the arguments that have been raised against the ILCS, involving it having the mortality question embedded in a larger survey, I'm not sure if the ability to produce death certificates in L12 & L2 necessarily applies here.
Thus, I view these results as being like what Les Roberts said about L1 and L2. They help show that the mortality is huge. I say this because I hope people won't get fixated on precise numbers. I take L2 and the ORB survey/poll as suggesting that the number of dead is in the many hundreds of thousands. Whether it is >1,000,000 or <1,000,000 I don't really know and may never know. There is just too much noise in the chaos of Iraq to get precise numbers.
I know that in surveys in the relatively placid US on such issues as tobacco or substance use you can get a fair spread between surveys, well beyond the CIs of the respective surveys, probably due to nonsampling error. Surely we'd find the same in Iraq.
So, now we know that there is much evidence that the mortality is enormous. We also have increased evidence, as some of us have argued for a long time, that passive surveillance systems, like IBC and the alleged Ministry of Health and US military figures, when not simply politically manipulated (to be clear, I'm referring to the latter two sources, NOT IBC here, in terms of the numbers), are missing the vast majority of deaths.
Again, I hope we can avoid the fetishism of the number as we absorb this new, horrifying, result. To avoid the fetishism of the number is also to acknowledge that there may be some odd results that seem counterintuitive, like the car bomb figure. These results may either show that our intuitions are wrong or that there is more noise in the figures than that reflected by CIs. At this point, we simply don't, and probably can't know.
Posted by: Stephen Soldz | September 14, 2007 1:56 PM
Ahh the troll JC trying to shift the metric to "media reported deaths". Laughable.
Posted by: Mondo | September 14, 2007 2:04 PM
Kevin, I think that asking someone if they favor attacks on coalition forces is already delving pretty deeply into dangerous territory. If for some reason a question about death in the family is regarded as even more sensitive, so sensitive that nobody has dared to ask it outside the Lancet surveys and this poll, then that in itself is a fact worthy of front page notice in the mainstream press--"Iraq is so dangerous pollsters don't dare ask about deaths in the family."
As for what Iraqi sources say, I know there was one claim by an Iraqi group of over 30,000 violent deaths before the end of 2003, and another claim in 2005 of over 100,000 violent deaths, both much higher than IBC figures for the same periods. These claims were mostly ignored. I don't know if insurgent websites provide enough detail for someone to make any sort of estimate on the level of violence.
On wounded to dead ratios, wouldn't people be more likely to forget a wound in the family if the family member recovered? I think this issue about surprising levels of forgetfulness came up in one of the earlier Lancet threads. But I'm dipping into technical issues now, which I don't know much about. (Not that I know much about anything in this area, for that matter.)
Posted by: Donald Johnson | September 14, 2007 2:38 PM
I am not asking this because I am trolling; I simply do not know.
Is this how casualty rates are normally assessed? It seems like there are so many ways that inaccurate numbers could be reached this way - basically you're just asking how many people are dead? I guess counting bodies is impractical... But doing it as a survey seems really flimsy.
Posted by: Marcus Ranum | September 14, 2007 3:03 PM
Donald, just a brief response while the English and South African rugby teams are singing their national anthems: it's probably a lot less dangerous to research whether Iraqis favour attacks on coalition forces rather than whether the Ministry of Health is cooking the morgue returns.
Beyond that I can only marvel at commenters whose argument runs: if ORB is sound then I have to change my views about the media, therefore ORB cannot be sound.
Posted by: Kevin Donoghue | September 14, 2007 3:10 PM
What's the surprise there? If you remember back to the precision missile and bomb attacks preceding the invasion, the Coalition of the Willing (Sick, eh?) destroyed principally civilian targets, not military targets. Rummy wiped out their phones, electrical power, water supply, transportation, and sewage system. The civil government collapsed, and criminals and warlords emerged as the new civil government.
When the military was quickly defeated, only the criminals and warlords remained in power.
The official story was that to help enable capitalism and democracy to arise spontaneously, the old order had to be destroyed first. Instead, we bombed them back to the stone age.
It helps to keep in mind that after the invasion was complete the great snark hunt began -- WMDs -- which occupied the attention of the coalition troops when they should have been rounding up all the munitions the Iraqi military left behind. Only when the last of the abandoned weaponry went into hiding did the snark hunt end. Rummy made the Iraqi insurgency the most lavishly armed insurgency in world history. We can thank Dickhead Don for supplying the Iraqis with the munitions they needed to produce their wonderful IEDs.
Posted by: Globle Warren Terrizm | September 14, 2007 3:14 PM
When there are more deaths than injuries, a likely answer is a policy of finishing off the wounded. When committing war crimes, eliminating witnesses is always a sound policy.
Posted by: Globle Warren Terrizm | September 14, 2007 3:17 PM
Onr things rings true in this survey - the high numbers of deaths by gunshot.
I remember a study published in Scientific American come years ago by two American academics that asserted the world's number one killer (other than disease or old age)was the AK47 assault rifle, the one of which arms-seller Samuel L. Jackson said in "Jackie Browne":
"If you have to kill everyone in the room , it must be an AK47"...
You cannot underestimate the lethality of modern automatic weapons, especially in enclosed spaces where bullets can be densely sprayed at the enemy or victim. The weapons in Iraq are probably ones stolen/ sold by the Saddam's disbanded army.
I do find the air attack deaths surprising high. However, we should not forget how ignorant we are of the true state of affairs in Iraq... the "media" do not have access to the whole of the country, being mostly confined to hotels or the Green zone, only getting around under with escorts.
Posted by: Toby | September 14, 2007 3:20 PM
Ditto with air strikes: 116,000 deaths (mostly unnoticed despite insurgent/militia/anti-occupation incentives to publicize them) and 132,000 injuries from air strikes. Given the nature of aerial bombardment, does this under-reporting or ratio make sense?
first point first: if do not accept the main result of a poll, do NOT bother with subresults!
or the other way round: imagine, if they had the bomb dead wrong by 50%! oops, next to zero effect on totals!
there are airstrikes in Iraq every day. after over 1000 days of war, the number of dead is rather low. and the majority of "air strike dead" might still have happened during the initial invasion.
http://www.alternet.org/waroniraq/62511/
A few other strange things from the detailed tables on ORB's website:
one more other strange things, from observing "iraq death toll critics": the majority of you has some serious lack in basic statistics!
1. It looks like they oversampled Baghdad relative to its share of the population. This could be important in pushing numbers up.
please, try to find out, what "oversampling" is. hint: it has something to do with SUBSAMPLES, that you are interested in. please assume, that the O.R.B. guys do HAVE some basic statistical understanding.
http://www.opinion.co.uk/who-we-are.aspx
2. Anbar is not included (as far as I can tell from the table listing provinces). Strange.
reading statistics, requires basic simple reading skills. in the article that Tim linked to, they explain that they ignored Anbar, because it was too dangerous!
http://www.opinion.co.uk/Newsroom_details.aspx?NewsId=78
4. They also interviewed a disproportionate share of non-Muslims (only 45% of respondents were "Muslim" overall and 28% percent in Baghdad -- the reset where orthodox, catholic, protestant, etc.).
please explain: what line in what table, did you take that information from?
http://www.opinion.co.uk/Documents/TABLES.pdf
5. The study suggests a lot of movement/displacement. External refugee flows probably mean that ORBs baseline population estimate for Iraq is too high, and internal displacement has probably increased the average size of remaining households as people flee to live with relatives. Together, this would alter the total number of estimated households in Iraq downward and increase the average household size -- suggesting the ORB estimate is probably skewed high.
wow. you really expect movement, death and displacement to bias numbers upwards?
instead of assuming, that those displaced might be the ones, who suffered the worst casualties? and will not be polled, because they are in a camp or abroad?
For all these reasons, it would be good to know more about the methodology here.
for all this reasons: i seriously doubt, that you are the right person, to challenge their methodology.
Posted by: sod | September 14, 2007 3:33 PM
Presumably the destruction of infrastructure means that people who have been wounded are more likely to die because they don't have timely access to medical care, clean water, etc.
Posted by: Holly Stick | September 14, 2007 3:40 PM
Hard to extrapolate this sample size to the entire population of Iraq, though, since all areas are not equal in violence?
Regardless of the actual numbers, even if this number overestimates the deaths 100 times over, it's still far too many.
Posted by: Inky | September 14, 2007 4:00 PM
Oh, so my kid is dead, but you had good intentions? Oh well, that's alright then. Carry on!
Posted by: SmellyTerror | September 14, 2007 4:09 PM
The real question, in my mind, is how many of those people would be dead if Saddam was still in power. I mean, "oppressed beats the hell out of dead" for most people. It's kind of scary when "we're from the government and we're here to help" becomes a question of whether it's 100,000 dead people more or less one way or another.
The whole thing is very depressing. If I weren't already a cynic and a nihilist, I'd have lost whatever faith I had in mankind.
Posted by: Marcus Ranum | September 14, 2007 4:23 PM
Marcus Ranum asked:
No. Usually, civilian casualty rates are never assessed.
Posted by: Robert | September 14, 2007 4:35 PM
Re Marcus in #22 and Robert in #31, I'd add that to the extent that they casualties are addressed, all of the various methods (census figures, surveys, exhumation of bodies) suffer from varying degrees of uncertainty. In the case of a failed state, where the infrastructure has effectively collapsed, that uncertainty can be huge, and it might be necessary to wait for the return of some semblance of stability to determine the scale of the disaster. One method for evaluating casualties after the fact is to analyze the structure of the population, and compare the relative sizes of groups by age and sex with what we would expect to find in the absence of a mortality crisis.
Kind of depressing, huh? It may be another twenty years before we can determine just how much of a mess we've made.
Of course, for the people who supported the war, I suppose that works out pretty well.
Regards, Bruce
Posted by: Bruce Sharp | September 14, 2007 5:11 PM
Somebody pointed out that the CIA Factbook has a comparative death rate page (updated 9/7/07) that lists the following:
Death rate (deaths/1,000 population) World - 8.37 U.S. - 8.26 Canada - 7.86 Iraq - 5.26
So according to the CIA Canada has a 49% higher death rate than Iraq. I'm not sure what methodology the CIA is using but I'm pretty sure it's bunk.
Posted by: joejoejoe | September 14, 2007 6:04 PM
Hi Joe --
This isn't necessarily wrong. Countries with more advanced health care systems and longer life expectancy overall can easily have crude death rates in excess of less developed countries. A larger portion of the population will be very old, and consequently, more likely to die.
Regards, Bruce
Posted by: Bruce Sharp | September 14, 2007 6:15 PM
Thanks Bruce. It seemed wrong at first but I now understand the dynamic at work.
Posted by: joejoejoe | September 14, 2007 7:20 PM
I wish we could get this level of discussion on the subject over at my post on the subject at Daily Kos, which you can find here. Stephen and joejoejoe have already been there, but you're all invited.
Posted by: Meteor Blades | September 14, 2007 9:22 PM
"Does this low ratio of wounded to deaths make sense? Car bombs injure a lot more people relative to those they kill (from blast, shrapnel, etc.)."
It is pretty hard to miss from inside 10 cm, and there are a whole lot of bodies dumped on the street with contact bullet wounds.
Posted by: Eli Rabett | September 14, 2007 9:56 PM
While the results are broadly consistent with the Lancet results - I think they are asking a different question.
The Lancet studies generally tried to COMPARE the mortality rates before and after the US began military operations - and indicate how many additional people died. This was an effort to gauge - not just how many people died from the obvious direct effects of war (people killed by gunfire - IED's - bombing raids - etc.), but also the people who died from the second order effects (diseases that were not treated well because the medical system was disrupted - people dying from overheating because the electrical system could not provide consistent power for air conditioning - etc.)
This study was asking individuals to make the assesment of whether someone in their family died because of the war - and pushes people toward only counting the "violent" deaths. I won't presume to know how Iraqis would answer this question if, for example, they had lost a family member to a disease that might have been curable if the US had not chosen to invade. Such an evaluation would require more knowledge of the exact text of the question (in the language it was asked) - and deep knowledge of Iraqi culture.
However, under the (I think plausible) notion that this survey ended up counting only the "direct" effect deaths (or even only some portion of the second order effect deaths), the Lancet results might actually have UNDER-estimated the Iraqi death rates. (NOT, I hasten to add, because of any pernicious intent on the part of the researchers - nor from any flaw in their method. Just because the Lancet studies could not be extremely precise, and were bound to be off in some direction.)
Posted by: lornix | September 14, 2007 10:50 PM
It's a survey, it must be correct! Right, Lambert?
Anyways, I have a brilliant idea. Why doesn't ORB employ the same methodology and figure out how many Americans have died from the war in Iraq? We know the exact figure for that, let's see how well their methodology does in a much more stable environment.
Of course, we'll just gloss over Iraqi culture where people may be considered the member of more than one household at any time within a span of four years. Not to mention that three provinces were not sampled at all. Etc. Etc.
Lambert doesn't care about such problems, so why should I?
Posted by: Seixon | September 14, 2007 10:52 PM
In #12, Robert noted: "I'm not sure where they get the 4.05 million HH number."
I don't know where they got that number either, but it's fairly close to the figures in the 2004 UNDP ILC survey. In the tabular version (http://www.iq.undp.org/ILCS/PDF/Tabulation%20Report%20-%20Englilsh.pdf), on p. 22, the report shows a figure of 4.252 million as the total number of all households. It also shows a mean of 6.4 members per household, and a median of 6 members. There might be more precise numbers buried elsewhere in the report, but it's a couple hundred pages of stuff, and I'm not motivated enough to wade through all of it.
Regards, Bruce
Posted by: Bruce Sharp | September 14, 2007 11:07 PM
Running the numbers, according to ORB's survey, there's about 27 million Iraqis currently living in Iraq. Which is funny, because supposedly a few million Iraqis have fled Iraq since the war. The population of Iraq according to Lancet and the UN back in 2004 was exactly that, 27 million.
Using ORB's data, approximately a minimum of 5 million Iraqis fled their homes, with a maximum of 11 million, with an average of about 8 million.
So, according to ORB's numbers, 27 million Iraqis currently live in Iraq, with at around 4 million having fled the country, even though there were only 27 million Iraqis in 2004 according to the UN.
Something doesn't add up here, for those who care.
Posted by: Seixon | September 14, 2007 11:27 PM
yes Seixon, and I'm sure that those who fled their homes had experienced much lower death rates than those who stayed. I'm sure they only fled to Syria because it has more KFC outlets.
Posted by: SG | September 14, 2007 11:40 PM
Seixon,
A two to three percent growth rate is not uncommon in third world countries. Iraq was estimated at 2.6% in 2002.
https://www.cia.gov/library/publications/the-world-factbook/fields/2002.html
Over more than four years, that does give you a positive gain of over three million. So obviously it is possible to have 3-4 million people leave the country without causing the population to drop significantly; there's nothing obviously incorrect about the numbers you're reporting, despite your claims otherwise. If you care.
Posted by: gz | September 15, 2007 2:10 AM
gz,
"A two to three percent growth rate is not uncommon in third world countries."
I see. So I guess since Iraq is just like all other third world countries, all other third world countries have the same displacement rate as Iraq? The same level of violence?
You're comparing apples to a maggot-infested apple. If you care.
SG,
So in response to a valid criticism of Lambert's complete disregard for critical thinking, you make some kind of joke to deflect, smearing me as someone who doesn't know why Iraqis would be fleeing their country.
Hey, whatever you have to do to protect the Emperor from the fact that he has no clothes.
Let's not even get started on the fact that the poll claims more people have died than have been injured by violence, which would be quite an amazing feat in the history of military conflict.
Let's just put our brain aside and nod like dumb sheep to whatever a polling firm releases, and whatever Tim Lambert tells us the Truth is.
All this has a name. It's called dishonesty.
Posted by: Seixon | September 15, 2007 2:52 AM
tsk, Seixon, I am not smearing anyone. (Whatever that consists of in comments to a blog).
I am merely observing that the displaced population of Iraq would likely have had a higher death rate than those remaining, and including them in the denominator won't necessarily artificially push up death numbers (as you imply) since their higher death rate has been omitted from the numerator. If anything it would lower death rates, since the denominator of the CMR calculation has been inflated by about 4% but the numerator has been calculated assuming that 4% had a lower death rate than was probably the case.
As a finite population correction in the sample statistics it might serve to reduce the variance a little, artificially increasing significance of the finding, but that's unlikely to have much effect for a population change from 27 million to 24 million.
Before citing the ratio of deaths to injuries as a concern, can you answer the point repeatedly raised about this, that people don't get to go to hospital in a war-torn country? Also you might like to consider this quote from the BMJ:
which finishes with the conclusion "a high ratio of killed to wounded suggests a war crime". (This paper gives the minimum ratio of injuries to deaths as 0).
Posted by: SG | September 15, 2007 3:49 AM
For the record, Seixon, nobody claims 4 million Iraqis "hav[e] fled the country." The generally accepted UN figure is 2.35-2.5 million, with another 2 million to 2.2 estimated to be internally displaced. Since there are no large refugee camps inside Iraq, presumably most have been displaced to other households. How many of them may have been surveyed, and how many may have lost members of their families would have been an interesting extra question for ORB to ask. I would think that displaced Iraqis might include comparatively high numbers of violently deceased members in their families.
I am not a statistician, or any kind of scientist, so my comments here I'm sure will be taken at lesser value than those with credentials. But I think some legitimate questions have been raised about the study, and I would be eager to know ORB's responses. But I am not willing to toss the poll out simply because some people consider the numbers far-fetched.
We all know that polling operations can do some awfully sloppy work. I've seen plenty of that myself, including from organizations like Gallup, and I think that should be taken into account in this case.
My chief concerns fall into two categories: respondent comprehension and truthfulness; sampling, methodology and extrapolation.
As we know from polls in the United States, respondents sometimes lie, as when pollsters ask white citizens whether they would vote for a black candidate. Why Iraqis, especially large numbers of Iraqis, answering such a poll might choose to lie I do not know. There might be myriad reasons. Likewise, they may have taken the question to mean something else than it does. I have not seen the Arabic version (my stepdaughter can translate for me), so I don't know if it might be misinterpreted in some way that would inflate the Yes answers.
Was Baghdad oversampled? I can't agree with Stephen that the failure to ask for death certificates does not harm the results. Because it definitely is a drawback to anyone trying to defend ORB's numbers in the non-scientific world.
I would certainly like to see someone with more brains and relevant experience than I have to ask these questions directly of ORB and then write them in a way that is understandable to a general audience.
Posted by: Meteor Blades | September 15, 2007 4:16 AM
I can't agree with Stephen that the failure to ask for death certificates does not harm the results. Because it definitely is a drawback to anyone trying to defend ORB's numbers in the non-scientific world.
Lancet did ask for death certificates. the "sceptics" simply assume, that the doctors faked those results.
there is NO way you can persuade persons, who deny to be persuaded. they are entitled to a math of their own. sigh.
I see. So I guess since Iraq is just like all other third world countries, all other third world countries have the same displacement rate as Iraq? The same level of violence?
do you have ANY evidence, that (natural) iraqi population growth is significantly down?
do you understand, that those who fled will have taken MORE casualties?
Posted by: sod | September 15, 2007 4:24 AM
sod, I know the people who did the studies that Lancet published obtained death certificates as part of their verification effort. My point was that the ORB surveyors did not do so, and that this "failure" will be just one more item used to hammer their results. Not that they wouldn't be hammered anyway, of course.
Posted by: Meteor Blades | September 15, 2007 7:01 AM
Oooh! Evolution at work!
[Typically in counterinsurgencies and civil wars there are at least 2-3 injured for every death.]
veterans of these debates will remember that in the past this factoid (which has little or no actual support from evidence - it is in fact mainly drawn from miniature wargaming rulebooks) referred to battlefield troops. Clearly this was a counter-adaptive quality for a factoid to have in the context of the Lancet debates, as it was rather egregiously irrelevant. Now we are seeing a new, mutated version of the factoid, where it refers to "counterinsurgencies and civil wars". It still has the fatal structural weakness of having no citations at all, but I suspect that the new version may proliferate a bit better. Gosh isn't evolution interesting.
The factoid is untrue by the way - when sectarian death-squads take someone out and murder them with a drill to the head, it is unusual for three other people to be caught in the cross-drilling.
Posted by: dsquared | September 15, 2007 8:06 AM
seixon says:
"Of course, we'll just gloss over Iraqi culture where people may be considered the member of more than one household at any time within a span of four years."
The polling firm said:
"Please note that I mean those who were actually living under your roof."
You can say that this qualification in the question might still cause confusion--you'd likely be wrong--but it is certainly not "gloss[ing] over" Iraqi culture. Claiming that it is would would be, well, let me use your words: "It's called dishonesty."
Posted by: Boris | September 15, 2007 9:56 AM
Looking at the earlier John Hopkins/Lancet study, one figure I jiggered out of the data is a combatant:noncombatant ratio inferred from the male:female ratio of 10:1.
A disparity that high beyond normal male:female distribution has to happen for a reason. I found it strange that the study authors didn't comment on the obvious disparity in male:female casualties?
From an earlier Washington Post article, dated Oct. 2006 .. http://www.washingtonpost.com/wp-dyn/content/article/2006/10/10/AR2006101001442_2.html
"....Of the 629 deaths reported, 87 percent occurred after the invasion. A little more than 75 percent of the dead were men, with a greater male preponderance after the invasion. For violent post-invasion deaths, the male-to-female ratio was 10-to-1, with most victims between 15 and 44 years old...."
The casualty ratio of men:women is 10:1 (violent deaths, not all deaths). The data strongly suggests to me that the higher probability of male mortality was b/c the bulk of them were in fact direct participants in violent situations, i.e. combatants. Unless I were to factor some kind of "honor code" proscribing the killing of women & children, I'll impute the male:female ratio as indicative of the larger (worldwide) cultural bias of male involvement in civil strife without any kind of "honor" discount.
Again, from the Washington Post article:
"...Of the total 655,000 estimated "excess deaths," 601,000 resulted from violence and the rest from disease and other causes, according to the study."
So let's do some accounting:
1) 55k / 655k = 8% chance of dying from disease (male & female) or 92% chance due to violence.
2) If 90% of violent male deaths died as combatants (10:1 ratio males:females violent deaths), we can infer that culturally the lower death rate for women is significant here, b/c they were likely noncombatants. We can use that ratio for all combatant:noncombatant deaths, then.
3) 655k * .75 = 491k male dead (75% total dead are male) 4) 491k * .92 = 451k males died from violence (not from 8% chance of disease) 5) 451k * .90 = 405k males died as combatants 6) 405k / 655k = 62% of the dead were combatants
That accounting suggests 62% of the total deaths (& 82% of male deaths) were combatant or those who were somehow participating in violence during JH's 3-year period (2003-2006).
During Saddam's reign - especially since the Gulf War in 1991 - the average Iraqi mortality numbers due to murder & disease was around 60k/year, 30k/yr due to murder & another 30k/yr. due to disease (in large part, child mortality).
There were also the NGO & an ABC-sponsored surveys that were contemporaneous with John Hopkins' surveys & used many more clusters. Those found a much lower overall death rate. Both were well above Iraq Body Count's (IBC) figures, indicating under-reporting problems with mere media surveys.
Hope that helps,
/lee
Posted by: leebert | September 15, 2007 11:25 AM
I've posted here a comparative chart and spreadsheet with the four surveys and the contemporaneous IBC body counts. The ORB survey estimate of total deaths by violence(1.22m) is within a whisker of Lancet 2 updated according to the IBC timeline (1.16m), assuming a constant IBC undercount.
Posted by: James Wimberley | September 15, 2007 12:03 PM
Leebert: "the bulk of them were in fact direct participants in violent situations, i.e. combatants." Like the Moslem men and boys at Srebenica.
Posted by: James Wimberley | September 15, 2007 12:08 PM
Leebert:
Could you tell me something?
If an armed group goes into a village and kills all the men and rapes all the women (not an unusual happenstance in terrorist insurgencies), what would be the male/female mortality ratio from that incident?
Posted by: Robin Levett | September 15, 2007 1:36 PM
leebert: "I found it strange that the study authors didn't comment on the obvious disparity in male:female casualties?"
How did you deduce that they didn't? Not by reading the study, that's for sure.
Posted by: Kevin Donoghue | September 15, 2007 2:06 PM
Stephen Soldz's remarks seem on-target to me. What this poll seems to show (unless there is a considerable amount of lying going on, or spectacular incompetence in sampling) is that the death toll is somewhere in the neighborhood of one million, give or take a few hundred thousand. You can argue it's less than a million if you think the number of households is less than 4 million, or you can claim that the refugees outside the country have a lower death rate or both, but you're not going to get the death toll all the way down to IBC levels that way.
So the fallback position is that either the respondents lied, the pollsters lied, or they did a really amazingly poor job picking out their sample.
Posted by: Donald Johnson | September 15, 2007 2:36 PM
veterans of these debates will remember that in the past this factoid (which has little or no actual support from evidence - it is in fact mainly drawn from miniature wargaming rulebooks) referred to battlefield troops.
it is not entirely tabletop based :)
the number i remember from army times, was between 3 and 6 to 1, and there were more detailed estimates about the seriousness of wounds and the casualty ratios for battalions performing different missions. i am rather sure those numbers can be verified in some older military books.
but those numbers are military wound/death ratios, in conventional battle. they are easy to track, because all soldiers wounded will be handled by the military medical system, as will be the dead.
the situation for civilians, in a civil war situation is completely different. the number given in one of the article linked to above (between 2 and 10 to 1, perhaps even below 1 to 1) sound absolutely right to me.
most factors for a different death to wounded ratio have been named already: shooting bound men, huge car bombs, failing medical system, people not remembering wounds, but deads.
Posted by: sod | September 15, 2007 2:44 PM
Meteor Blades,
"For the record, Seixon, nobody claims 4 million Iraqis "hav[e] fled the country.""
That's what ORB's data suggests, which was exactly my point. It's ludicrous, but none of you anti-war liberals would ever be critical of a single damn study that comes up with what you want. Don't let me stop you from BELIEVING in things that on their face are ridiculous. That is your wont, which you display quite well on DailyKos on a daily basis.
Posted by: Seixon | September 15, 2007 3:39 PM
Lee,