Lancet

by on October 14, 2006 at 6:52 am in Data Source, Political Science | Permalink

Left-wing bloggers, such as CrookedTimber, Brad DeLong, and Tim Lambert, are supporting the claim of about 600,000 extra deaths in Iraq.  Jane Galt (scroll down for a few posts) and Steve Sailer raise some concerns.

I am a bit skeptical, but in any case the sheer number of deaths is being overdebated.  Steve Sailer notes: "The violent death toll in the third year of the
war is more than triple what it was in the first year."  That to me is
the more telling estimate.   

A very high deaths total, taken alone, suggests (but does not prove) that the Iraqis were ready to start killing each other in great numbers the minute Saddam went away.  The stronger that propensity, the less contingent it was upon the U.S. invasion, and the more likely it would have happened anyway, sooner or later.  In that scenario the war greatly accelerated deaths.  But short of giving Iraq an eternal dictator, that genie was already in the bottle.

If the deaths are low at first but rising over time, it is more likely that a peaceful transition might have been possible, either through better postwar planning or by leaving Saddam in power and letting Iraqi events take some other course.  That could make Bush policies look worse, not better.  Tim Lambert, in one post, hints that the rate of change of deaths is an important variable but he does not develop this idea.

We all know that the political world judges Iraq by the absolute badness of what is going on (which means Bush critics find a higher number to fit their priors), but that is an incorrect standard.  We should judge the marginal product of U.S. action, relative to what else could have happened.  (North Korea, and the UN response, will give us one data point from another setting.)  In that latter and more accurate notion of a cost-benefit test, U.S. actions probably appear worst when deaths are rising over time, and hitting very high levels in the future. 

Of course the rate of change of deaths is not exactly the proper variable.  Ideally we would like some measure of the contingency of eventual total deaths, relative to policy.  I am not sure what other proxies for that we might have.

Addendum: Let me put my comment up here on the front page: "Many of you are misreading the post by focusing only on the first case
of "bottled up killing," which is presented as only one of two
scenarios.  Reread that if deaths are rising over time and possibly
contingent — and yes I do say this is the relevant and uncontroversial
fact — this suggests a very negative evaluation of Bush policies." 

I don’t want to take the bait on why I am skeptical, the whole argument is that possible skepticism doesn’t have that much import once we consider the broader context of rising deaths and the possible contingency of those deaths. 

SomeCallMeTim October 14, 2006 at 8:57 am

A very high deaths total, taken alone, suggests (but does not prove) that the Iraqis were ready to start killing each other in great numbers the minute Saddam went away.

That needs some justification. Or at least an acknowledgement that the claim weakens the case for invasion: Hussein wasn’t going to live for ever, and apparently the Ba’athist regime wasn’t sufficiently stable to sustain a nuclear program past his death. So choose which strut of the justification you want to kick out from under the case for war: Hussein was a threat to the US, or there was a humanitarian reason to do this.

Once that’s done, we can address the remaining strut.

bjak October 14, 2006 at 9:23 am

They were all going to die anyway is a great rationale for a war. We should do a cost benenfit analysis of all the wars that are going to happen anyway, calculate the benefit of preemptive action, and then start those wars now to save those hypothetical future lives. Among the many other problems with this argument is the tone of false precision.

Tim Lambert October 14, 2006 at 12:04 pm

Your “a bit skeptical” link goes to a case where there was an overestimate because a convenience sample was used. I don’t see how this is a reason to be skeptical of the Lancet study which used a random sample.

And I agree with Matt about the Galt posts. I don’t think her anecdote about New Coke is a cogent criticism of the Lancet study.

Kevin Nowell October 14, 2006 at 12:42 pm

I don’t see why the actual death count should matter so much. If the war was immoral, it was immoral on a priori grounds.

If you are doing a cost/benefit analysis it makes no sense to do it from the Iraqi perspective as they were not the ones that made the decision. If doing it from the American perspective, as proper, then the 600,000 deaths in Iraq don’t come into play on either side of the ledger.

Jor October 14, 2006 at 1:31 pm

What exactly are Galt’s technical concerns with the study? If anything, it seems to be completely her opinion. Maybe people are biasing their interpretations with prior beliefs — but at least some people are using evidence, while others are just name-calling.
Sailer’s raises some points, but seems to conclude, he could believe 300K deaths. Not exactly a stinging refutation on any methodological points with the study.

Ennis October 14, 2006 at 1:57 pm

A very high deaths total, taken alone, suggests (but does not prove) that the Iraqis were ready to start killing each other in great numbers the minute Saddam went away.

I don’t see your reasoning here either. Is the holocaust evidence that the Germans had a bottled up propensity to murder Jews, Gypsies, Slavs, etc? Goldhagen says yes, but most other scholars emphasize the contingency of the killing, despite the mass scale.

How about Cambodia? Or Rwanda, where the level of mass involvement was high? Experts on each of these incidents disagree.

Heck – how about the US civil war? Is that evidence that northerners and southerners would have slaughtered each other in large numbers at some point, even if secession had been avoided and slavery had been peacefully phased out?

Tyler, think hard about the mechanisms involved in this killing right now. It’s not about every Iraqi turning around and killing his neighbor, it’s about a situation where ethnic militias have developed as part of a cycle of escalation. It took a while for these organizations to come about and be mobilized. All of that could have been prevented.

You’re seeing propensity, I’m seeing response to specific incentives created by the occupation.

I can elucidate further, with more specifics, but it’s probably better for me to make my case off-line.

Anderson October 14, 2006 at 2:22 pm

Like Lambert, I can’t function at the Cowanesque level required to see how the African-statistics link provides any basis for being skeptical about the Lancet study, other than perhaps a general “people get stats wrong sometimes” skepticism.

I was looking forward to seeing MR weigh in, since presumably y’all are adept with statistics and your opinion would carry some weight.

Tyler Cowen October 14, 2006 at 2:26 pm

Many of you are misreading the post by focusing only on the first case of “bottled up killing,” which is presented as only one of two scenarios. Reread that if deaths are rising over time and possibly contingent — and yes I do say this is the relevant and uncontroversial fact — this suggests a very negative evaluation of Bush policies.

Jason Voorhees October 14, 2006 at 2:59 pm

What is accounting for the huge discreprancy between the Lancet estimates and the Brookings Institute’s estimates (see here).

“On the other hand, Michael O’Hanlon of the Brookings Institution, which tracks statistics in its Iraq Index, said: “I do not believe the new numbers. I think they’re way off.”

The Brooking Index, relying on the UN (which gets figures from the Iraqi health ministry) and the Iraq Body Count (IBC), estimates the civilian death toll at about 62,000.”

Anyone familiar enough with the two methodologies and data to comment about why such a vast disparity exists?

Kevin Donoghue October 14, 2006 at 3:28 pm

Jason, your own link to the BBC provides a reasonably good summary. IBC acknowledges that it only picks up a fraction of violent deaths and the Iraqi health ministry figures are woefully incomplete. As Les Roberts says: “There have to be ~300 deaths per day from natural cause even if Iraq was the healthiest 26 million people in the world. Where are those bodies? When the MOH [ministry of health] in Iraq is perhaps recording 10% of them, why should they be doing better with politically charged violent deaths. Yes, I think almost nothing is getting reported outside of Baghdad where things are worse.”

joan October 14, 2006 at 6:55 pm

Iraq Body Count (IBC) has no one on the ground in Iraq but collects information from cofirmed news reports ( two or more sources), so it is an significant undercount. As the news media retreated to Baghdad the undercounting probably is even more of a problem. What ever you think the current count is, it is large enough to merit concern. My fear is that it could grow to levels seen in Cambodia if we leave or could go on like this for decades like Viet Nam if we stay.

Barkley Rosser October 14, 2006 at 9:01 pm

Jason Voorhees,

Besides the fact that Iraq Body Count only counts deaths reported in
newspapers, the other reason why official numbers may be off is that
apparently there has been a severe breakdown in communication between
local ministries and the central one in Baghdad. Thus, many death
certificates are handed out locally, but never recorded in Baghdad.

K.,

A similar point can explain the problem with the media. There is little
media that is really reporting from around Iraq, with all too many reporters
of one sort or another never leaving the Green Zone in Baghdad. More to
the point, some of the most dangerous places are essentially never visited
by any foreign media anymore. They are too dangerous.

Biomed Tim,

The sample is much larger than any other that have been the basis of studies
in Iraq and is very widely based geographically. This of course does not
guarantee that it may not still be biased upwards to some extent.

Jason Voorhees October 14, 2006 at 11:20 pm

I understand that the Iraqi Body Count is an undercount. But it’s a 10th of the size of the Lancet study. Is that a reasonable undercount? Have demographers found similar magnitudes of undercounting before with that type of data? I’m just having trouble understanding how it’d be possible for the differences in these two data sources to exist, even with the undercounts. It’s a completely useless method of collecting data if the real number is 600,000 and body counts only can get 10% of that. Is this the first time those counts have been called into question like this? Not being a demographer, this is new to me.

Douglas Knight October 15, 2006 at 1:47 am

Biomed Tim,
if 95% confidence is wrong, please correct the arithmetic error. Those concerns about selection bias you link to flatly contradict the article. Have you read it? Are you saying the authors are lying?

Jason Vorhees,
The article says “Aside from Bosnia, we can find no conflict situation where passive surveillance recorded more than 20% of the deaths measured by population-based methods.” It would be nice if they expanded on that claim, but aparently 10% is a reasonable undercount.

Taeyoung October 15, 2006 at 8:49 am

A very high deaths total, taken alone, suggests (but does not prove) that the Iraqis were ready to start killing each other in great numbers the minute Saddam went away.

Probably, although outside instigation has probably played a role (e.g. the golden mosque bombing) — that has to be factored in anyhow.

Reread that if deaths are rising over time and possibly contingent — and yes I do say this is the relevant and uncontroversial fact — this suggests a very negative evaluation of Bush policies.”

Fair enough — the question then is what policies would have avoided that result? The scenario you seem to be outlining is that (a) Iraqis wanted to kill each other => high death total, BUT (b) while Saddam ruled with an iron fist, suppressing all internecine slaughter, they couldn’t do so, AND (c) immediately after we crushed Saddam’s government, the Iraqis hesitated to start killing each other. Why?

I can think of a number of reasons, but the simplest is that they thought we would come down on them Saddam-style if they started it up, and when we didn’t come in with the mass executions and reprisals, they took heart and went to the slaughter (at least in some areas) with gusto. And that might have worked — the British put that kind of brutality to work in pacifying the entire Raj with a smaller force than we have in Iraq. But that wasn’t ever in the cards. Then the complaints about neo-imperialism really would have been correct.

That said, I think the really key factor here is that elements within Iraq, possibly aided by outside elements (e.g. the Persians, Zarqawi, etc.) managed to outplay the occupation by triggering internecine slaughter with high-profile provocations like, again, the golden mosque bombing. It’s not been a straight-line development — the population is pretty mixed, after all, which makes stark Sunni-vs.-Shiite conflicts harder to bring about — but cumulative.

And that said, I think the Lancet numbers are suspect, largely for the reasons Megan McArdle has already pointed out. Marcus’ point about people evidently fleeing toward the killing fields is also persuasive.

bernard Yomtov October 15, 2006 at 11:51 am

We should judge the marginal product of U.S. action, relative to what else could have happened.

Yes. But “what else could have happened” should include things the US could have done differently post-invasion. I think you meant to consider that, but I wonder whether everyone making these sorts of “they’d kill each other anyway” arguments does.

And if Iraq was headed for this kind of slaughter anyway, what exactly was accomplished by the invasion?

Marcus October 15, 2006 at 3:57 pm

“More than 50,000 exiled Iraqis returned from neighboring countries last year in the hope that calm might return after the country’s first post-war elections in January 2005. That number has fallen to 1,000 this year. “Far more are leaving,† Ron Redmond, chief spokesman for the United Nations High Commissioner for Refugees (UNHCR), told a news briefing in Geneva.”

The phrase “has fallen” implies greater rates of return before the present period. Recollect you are reading a UNHCR document, the UN being no fan of the invasion to put it mildly. Just as you would read an Administration report under one light, you should read a UN document under another. During most of the period covered by the Lancet study, the flow of refugees has been net inward, not outward. Even the UNHCR reports they have been closing refugee camps until recently.

K October 15, 2006 at 4:43 pm

Valuethinker, et al: Yes I understood that reporting has been dangerous and perhaps suicidal in these areas. But that does not excuse maintaining coverage that gives no good information.

If there is such an utter difference between media coverage and fact then exactly why are war supporters mistaken? Or is it ‘cut and run’ that is makes no sense? Lacking facts people tend to resort to political and moral beliefs i.e. the other party is wrong, anything is better than X.

From history (I’ll assume some is useful for my arguments) we see that news during war is usually unreliable. We also would observe that wars are not linear – long lulls occur while resources are gathered and strategies are reviewed; then suddenly all hell breaks loose and 100,000 die in a day or week. Lulls also occur when the attackers are somewhat exhausted and overextended – is that peace breaking out?

The Pacific in WW2 saw roughly 3.5 years of conflict. But the last 6 months were probably the bloodiest even though Japan had clearly lost. So increasing civilian or military casualties indicates what? Well in 1945 they indicated we were firebombing, then atom bombing, Japan. But the rising tide of deaths did not indicate stablity or peace was not near.

Which brings me back to this Lancet flap. The methodology, to my intermediate understandings, looks pretty good. The data I can’t know anything about. If I can’t believe almost everything else reported about this war then what weight is to be given this?

Taeyoung October 15, 2006 at 7:46 pm

Re: Yomtov

And if Iraq was headed for this kind of slaughter anyway, what exactly was accomplished by the invasion?

We destroyed the Baathist government in Iraq and took all their uranium. Hopefully we’ve captured or killed all or most of the people who were in command to, such that if they hid significant weapons in their territory, everyday terrorists will have nearly as much trouble as we have finding it.

Those were our primary war aims after all, even if part of the justification underlying those aims was that the Iraqi people were suffering under grinding oppression. The primary justification, of course, was that we didn’t want to wait until Saddam had a bomb (until the threat was “imminent”) before taking him seriously — then we’d be in the situation we were in with North Korea in 1994, where all we can hope to do is buy him off for a little while longer. Or, indeed, in the situation we’re in with North Korea today, where we really can’t do anything or Kim Jong Il will raze Seoul and try to nuke Tokyo.

That’s . . . no longer an issue with Iraq. Iraq poses about as much threat to us now as Somalia did in 1992. Maybe 14 years hence Islamists will successfully take power in Iraq too.

That aside, humanitarian concerns do underpin the continuing occupation, so while I would oppose withdrawal, if it would improve the lot of the average Iraqi (e.g. if recruitment to internecine slaughter within Iraq would fall because there’s no longer the shame of Americans rubbing Iraq’s crushing military defeat in their face), it might be a reasonable policy. It’s not like there’s a major downside for us, aside from the PR boost it gives our enemies. This was a war of choice, after all. However, I don’t see any reason to believe the situation would improve significantly with our departure.

Re: Klein

The methodology involves extrapolating 666,000 estimated deaths from 547 reported ones.

That’s what pollsters do all the time — and why the Lancet’s margin of error is ginormous. The question is whether the 547 data points they’re extrapolating from can reliably be extrapolated to the rest of the country (i.e. are they, so to speak, a random sample).

Mark October 15, 2006 at 9:19 pm

…we would expect to see more people fleeing the country than returning. But the flow of refugees during the bulk of the period covered by the study has been inward, not outward. The UNHCR has been closing refugee camps. And in 2005, the holy sites of Karbala and Najaf received approximately 12 million pilgrims, making them the most visited spots in the entire Muslim world, ahead of both Mecca and Medina.

Many of the refugees are Kurds who fled Iraq after 1988. Iraq is a diverse country and just using aggregate refugee flows as an indicator of the political situation is not going to give an accurate picture. They are now returning because Kurdistan is relatively safe compared to the rest of the country and there is less uncertainty than when the no-fly zone was the only thing protecting Kurds from genocide. Karbala and Najaf may also be relatively safe but they are only the 7th and 8th largest cities in Iraq with a combined population of 790,000 or about 3% of Iraq’s population. By contrast, Baghdad — by 1997 numbers — has nearly 20% of Iraq’s population.

See http://www.citypopulation.de/Iraq.html for the population numbers.

Biomed Tim October 16, 2006 at 12:22 am

GT,

I misquoted when I said the sample size might be too small. Here is the actual quote:

Robert Blendon, director of the Harvard Program on Public Opinion and Health and Social Policy, said the number of deaths in the families interviewed — 547 in the post-invasion period versus 82 in a similar period before the invasion — was too few to extrapolate up to more than 600,000 deaths across the country.

Blendon is much more of an expert than I am so I thought it was worth mentioning. Perhaps other people better versed in statistics can expound.

Anderson October 16, 2006 at 9:11 am

Shorter Marcus: “I present no evidence of my own, but your evidence is suspect because it comes from a dubious source, not to be confused with Marcus, whom you should take on trust.”

Tom Myers October 16, 2006 at 9:29 am

I find the study basically convincing — the death rate may only have doubled, not more-than-tripled, but things have gotten pretty bad (and yes, I do blame Bush in significant part.) My response is one I don’t see much around the web, though: I think that we are entering an era of desktop/personal manufacturing, and if Bush’s “Big Bang” (as TPM Barnett calls it) doesn’t work, that will be an era in which losses of 655,000 will not seem large. Dictatorships breed terrorism, religious fanaticism makes it worse, and Moore’s Law, shifting from chips to machinery via projects like fablab.mit.edu and reprap.org, will enhance their productivity too.

K October 16, 2006 at 1:09 pm

Klein: Methodology refers to method. Your 547 refers to data.

Here the margin of error is enormous* and the study was probably intended for political purposes. I suspect there was fraud – but that would be from the actions of people and not the selection of clustering per se.

And my suspicions are only suspicions, not evidence.

Besides faking data there are other ways to get the results you want.

Simply start in a few damaged areas. Survey them honestly, then decide it is dangerous or needless to gather more data and stop.

Or select areas where many transients are living – they came from somewhere else and it is probably because in that ‘somewhere else’ people were being killed.

I don’t know how or why this study was made. Whether one believes it or not, subsequent actions depend not upon the numbers but upon core beliefs.

Some will say even one death is too many – in which case the entire earth should be disarmed, who would enforce that? Others say it is all Bush’s fault – the premise of the evil/idiot leader. Others say it depends – 55,000 is er, well, sort of OK, whereas 655,000 proves we should withdraw. Don’t forget Western civilization zealots or those awaiting the return of Jesus – they have other ideas.

I know! Congressional hearings should be held!

*Technically I am not sure a margin of error should be said to exist when it is stretched this far. But I took advanced statistics long ago and far away, and defer to those better trained and currently analyzing stuff like this.

Anonymous October 13, 2008 at 10:54 pm
Electric treadmill January 12, 2010 at 5:21 pm

very high deaths total, taken alone, suggests (but does not prove) that the Iraqis were ready to start killing each other in great numbers the minute Saddam went away. The stronger that propensity, the less contingent it was upon the U.S. invasion, and the more likely it would have happened anyway, sooner or later. In that scenario the war greatly accelerated deaths. But short of giving Iraq an eternal dictator, that genie was already in the bottle.Galileo thermometers
Huggable hanger
nose hair trimmers
Scaffold boards

Comments on this entry are closed.

Previous post:

Next post: