Logic Times

Comments

Page Two

The Civilian Casualty Fable

14: Paul Garner

December 6, 2005 12:07pm EST

Dear Sir;

As a retired Air Force Sergeant (1954-1975) with almost 5 years in that exotic garden paradise of Southeast Asia (Vietnam, Cambodia, Laos and Thailand) I want to congratulate you on making a truth so easy to understand.  Our troops and the motives of our government have been badmouthed by the Left since the end of World War Two.  The Left ignores the fact that our military has undergone a major change since 1946.  We have moved from the massive civilian conscript military force of WW2 to the highly trained and professional military force we currently have.  Although this training is most visually evident in out Regular forces, it is also present in our Guard and Reserve forces.  Our forces try very hard to avoid civilian causalities.  We do try to limit combat only to the enemy (enemy defined as the chap who is shooting at you with the intent of doing you grievous harm or making you dead) and we have succeeded beyond all reasonable expectations.    I only regret that there was not anybody like you to take the conflicting claims put forth by the hippie denizens in the Peoples Republic of Berkley and other locations about Vietnam and our troops there and work them into something that more reflected actual fact.  

As I approach my 70th year, I can honestly look back and say that I am proud to have served.  Even more so, I can say how proud I am of those who are serving now.  They are better than we were.   May the Good Lord bless and protect every one of them.

Paul Garner

US Air Force (Retired)

{Aslan: It is feedback such as this that makes Logic Times worthwhile.  Thank you for your service and the kind words.}

15: Andrew

December 7, 2005 12:51am EST

A brilliant piece of work, my hat is off to you.

When the "Iraqi Body Count" became big news recently, my blogging cohort and I questioned the peculiar numbers for some of the very same reasons you have here.  We did not, however, pursue them in such a thorough and revealing matter: Iraq Body Count Speaks

The numbers you have deduced are utterly incredible and illuminating, and at the same time not that surprising at all.  The body counts purveyed by the anti-war crowd were always suspect in my mind, and this brilliant analysis of yours quite handily blows them out of the water.  

A small aside.  The one figure that always bugged me, and the one that you zeroed in on is the discrepancy between male and female deaths.  It seems utterly impossible that so large a discrepancy should exist if the deaths are truly "collateral."  Indeed, there's no doubt in my mind that this is the case, and that most of those "civilians" are in fact combatants.  However, other factors may have played a part in swaying the numbers, although how significant they are I cannot say.

As was touched on by a reader before me, I consider where the fighting is happening.  Will there be more civilian men than civilian women? Possibly.  If we're talking about roadside bombs and ambushes, more civilian men would probably be present on freeways and roads.  But then again, I cannot say where exactly the fighting happens.  Continuing, it seems more likely that a civilian male would be more likely to be mistaken for a terrorist.  Either way, this is irrelevant when countering the arguments that American soldiers are deliberately targeting civilians.

Again, I stand humbled by the excellence of your work.  

Andrew

Saskatchewan, Canada

{Aslan: Thank you, Andrew.  Your work, which I recommend all read, was far ahead of mine in questioning the statistical accuracy of the IBC study.  Nicely done.  As for the gender discrepancy, I would suggest your observations and the observations of others are accurate to the point of skewing the numbers slightly, not from 49.5% to 18.3%.

Your location in Canada reminds me of a humorous map I drew after the election in 2004 where I wanted to join forces with the wiser Canadian interior (our new country in maroon):

Let’s continue to work on this!}

16: Alistair Morley

February 24, 2006 8:24am EST

{Aslan: What follows is an exchange with a reader who is an expert statistician.}

Sir,

Just thought I'd write to commend you on your ongoing numerical analysis of Iraq.

It might interest you to know that myself and some colleagues independently ran exactly the same numbers in our free time, on both Iraqi Body Count and the Lancet piece, and came to exactly the same conclusion about the collateral damage. Good to know some people out there are doing rational analysis with the evidence. You've even identified the major assumptions about random target selection and exposure by category.

Actually, I think you can tighten your argument further. The criticisms of  your "assumption of uniform population exposure" are worth examining because both can be rejected/bounded by the same data. Consider:

# "Adult Females Civilians and children mistakenly selected less as targets than Adult Male Civilians"

 

Both Lancet and IBC data make it plain that a large proportion of all casualties are by "heavy" blast and fragmentation effects; these weapons effectively randomly select any _exposed_ target in the vicinity. Hence this counter-hypothesis is insufficient by itself to explain the massive observed gender discrepancy, (though with a bit of work it can be used to create an upper-end estimate on deaths-through-mistaken-targeting, which isn't very large)

 

Continued Top Right

16: Alistair Morley

February 24, 2006 8:24am EST

(continued)

# "Adult Females Civilians and children are _exposed_ less to danger ("they stay indoors")

 

Again, a neat way to check this counter-hypothesis can be found in the heavy weapons casualties. Look at a sample of death from terrorist car bombs. These randomly injure anyone exposed in the vicinity, male or female. Hence, a large enough sample of male and female casualties from just these type of attack will give you a good working estimate of the ratio of men-to-women "out on the street" during the day. Once you get that odds multiplier, it can be used to control for the overall male civilian deaths with more confidence.

As a final aside, I personally felt the Lancet piece itself was _relatively_ constrained and methodologically sound, and reported in good faith. However, the accompanying editorial and media exposure were classic examples of bias through selective under-reporting. As the media outlets should have or did know better, they qualify for the charge of either gross negligence or bad faith, take your pick.

Anyway, as a fellow Operations Researcher and fan of logic, may I wish you all the best with your ongoing work.

Alistair Morley

{Alistair:

Wow. There are compliments and then there are compliments, and yours, due to your expertise, ranks up at the top.  Thank you very much for the feedback. I am an amateur, but I have a pretty good nose for numbers and so value your comments.

Regarding your comment on the Lancet study, what is your opinion of the range offered: "We estimate that 98 000 more deaths than expected (8000-194 000) happened after the invasion outside of Falluja and far more if the outlier Falluja cluster is included."

I would like to post your comment, but seeing your disclaimer, I wanted to get your approval. If, given your profession, you prefer it not be posted, that is quite alright.

Thanks again.

Dan Hallagan

Logic Times}

Thank you for your reply; I'm glad you found it useful. I really don't think us "professionals" can claim much of a lead here; most good analysis doesn't require rocket science, insider info, or massive computers, but rigorous thinking and a healthy dose of skepticism. Talented amateurs can, and do, get there first.

Actually, your reply prompted me to go back and look over the Lancet article again. My take on the error ranges is that they are large, but what I'd expect given the clustering of the data. The forecasting technique is legitimate, but Fallujah is an outlier in so many ways, and anyway there are grounds to suspect fraud in the numbers here*. I too would want to see Fallujah removed from the data. The total quoted without Fallujah; 98,000 excess deaths, looks "reasonable" (Checked by independent if cruder method), but of these only a minority are from 'violence' and an even smaller fraction will be 'collateral damage by US forces' as it is commonly understood.

Which brings me to the main problem with the Lancet; the disingenuous breakdown/ misrepresentation of that total by various media. [This BBC article offers an excellent example of how to conflate "violent deaths"/"civilian casualties"/"excess deaths" totals for rhetorical effect. Incompetent reporting, or mendacity?]

One can easily dis-aggregate the Lancet to get a figure for post-war deaths supposedly caused by US action:

142 War and Post-War deaths [total]

 

90     less 52 from Fallujah

77     less 13 accidents**

21     less 56 natural causes

16     less 5 from the war itself (Mar/ Apr 03)

Of these remaining 16 deaths, we're critically not told how many are caused by coalition action (let alone how many of those were insurgents). Reading the article closely, its possible to see that between 5 and 8 of them were caused by US forces. Conversely, between 11 and 8 were caused by insurgents or criminal activity (12 in total were, but 1 was in the war itself and some might have been in Fallujah). Keeping Fallujah out, and re-scaling with these numbers gives:

Deaths caused by US military action in period May 03 to Oct 04:  ~7,100 (95% 15,700 /  5% 2,700)

There is no easy way to separate out the proportion of these which are insurgent deaths with the data provided (sample too small). A similar, or slightly higher number of deaths is apparently caused by criminals/insurgents, all of which would presumably be non-combatants.

However, you can see that although the 98,000 may not be a bad working figure for overall deaths, once one tries to drill down into the detailed sub-groups, the samples become far too small (just 5 deaths!). Hence despite the media hype, one really can't reliably use the Lancet figures to get "deaths by US military action" let alone "civilian deaths by US military action." To put these figures further in context, one can observe that Iraq has a normal mortality rate, all causes, of about 130,000 people per year in a population of about 24 million. So historically speaking, these figures are unprecedentedly low given the scale and intensity of the violence.

These are solely my personal comments, so I'd be more than happy to be posted by name, but please leave my other details off.

Best regards,

- Al

* The Juvenile-to-Everyone-Else death rate for Fallujah is too high even for a random bombardment of a city in peacetime, let alone selective bombardment of one that was mostly evacuated and occupied by insurgents at the time of the deaths (see deaths timeline). Given most deaths were by blast weapons, and juveniles would be more like to evacuate than women, its practically impossible this distortion could have happened by chance. One must suspect many of the respondents were lying to the interviewers; either the deaths did not occur, or were mis-attributed to US forces.

**One thing the Lancet researchers _don't_ pick up on is the unusual surge in the accidental death rate. A large rise in personal car ownership and use, and a boom in the construction / reconstruction trade, has occurred in Iraq since 2003. But saying "Iraqi's more likely to be killed by their new car than stray bullets", doesn't quite have quite the same ring in the media.