April 09, 2003
Why 'Friendly Fire' Deaths In Iraq Are So High

Of course, 'friendly fire' deaths in Iraq have not been high in absolute numbers, just a painfully high percentage of total coalition deaths.* This is not as surprising as it appears. Simple statistics can sometimes show odd results, if the disproportion is large enough. It seems to me that the high percentage of allied deaths from 'friendly fire' is in great part a direct result of the very success of coalition strategy and tactics.

Consider a couple of examples:

1. If allied forces kill 1000 people for each one lost to enemy fire, that's very good, about as good as anyone could reasonably expect. If they are 99.9% successful in avoiding 'friendly fire' deaths, that's also very good. But the combination of these two good things has the perverse result that fully half of their deaths will be from 'friendly fire'. Specific hypothetical numbers will show what I mean. Suppose that in a given week Iraqi soldiers kill 10 allied soldiers in battle. Assuming a 1000:1 ratio and 99.9% accuracy, that would mean that coalition forces would kill 10,000 people in the same week, 99.9% of whom are not allied soldiers -- say 99% Iraqi soldiers, 0.9% civilians, 0.1% allied soldiers. The result of our hypothetical case is that the Iraqis lose 9,900 soldiers and 90 civilians to Allied fire, while the Coalition loses 10 soldiers to Iraqi fire, 10 more to 'friendly fire'. (To simplify, I omit journalists and all those killed in ordinary accidents, collisions and such.) The very competence of coalition troops, as shown by (a) the huge disproportion in casualties and (b) the very high accuracy in targeting, produces a shockingly high percentage of deaths from 'friendly fire'. Since coalition forces take good care to determine and publish the cause of each allied death, 50% of relatives in our hypothetical case have the added misery of knowing that their loves ones died at the hands of their fellow soldiers.

2. The same argument works from the other side. Developing example 1 further, suppose Iraqi troops are so astonishingly incompetent that they not only kill very few people compared to their losses (that 1000:1 ratio again), they even kill just as many of their own soldiers as they do of ours. (This is probably too high, but I'm making an à fortiori argument. To simplify again, I omit from my calculations any civilians or journalists they may succeed in killing, whether intentionally or not, plus any soldiers shot by their officers for refusing to fight, successful suicide bombers, officers 'fragged' by those planning to desert, and so on.)

Continuing with the previous assumptions, that would mean that they lose 9,910 soldiers, 9,900 to coalition fire, only 10 to their own incompetence, despite the grossness of the incompetence hypothesized. Even if every death could be clearly assigned to its cause -- unlikely in this war --, only a tiny percentage of grieving Iraqi families would have the added misery of knowing that their loved ones had died at the hands of their fellow Iraqis. (Again, I'm counting only those killed in battle by misaimed weapons, not executed deserters and such.)

I don't see any solution to this statistical paradox. We certainly don't want to reduce the ratio of enemy deaths to allied deaths. The only possible improvement is a further increase in the accuracy of targeting. Of course, in this imperfect world it's very difficult to increase a number that's already at 99.9% of the theoretical maximum,† all the more so in the barely controlled chaos of modern war. The only sure way to get the number much closer to 100% is to add further checks and rechecks and crosschecks that take up precious time, time which allows even a sluggish and badly led enemy to react, and thereby causes the ratio of enemy deaths to friendly deaths to worsen. It is surely better to lose 20 men in battle, half to 'friendly fire', than to lose 40 or 50 or 100, only 1 or 2 or 5 of them to 'friendly fire'. Not that that would make the survivors feel much better.

- - - - - - - - - - - - - - - -

* 'Friendly fire' is not a good name, but it is concise, and more recognizable than 'blue-on-blue'. I will use it here, but only in quotation marks.

† I don't know the actual number, but it's certainly something in that general area.

Posted by Dr. Weevil at April 09, 2003 11:59 PM

I believe the common usage in the military is "fratricide".

Posted by: David Perron on April 10, 2003 11:43 AM

No, it is 'friendly fire', as in "Friendly fire ain't". I would use the word 'fratricide' if I was to brief someone who needed to be impressed by my erudition.


Posted by: J.M. Heinrichs on April 10, 2003 10:33 PM

I doubt that anyone at all familiar with military history is bothered by the friendly fire statistics at all. Back in WWI, it became infantry doctrine that if your artillery wasn't killing a few of your own troops in an attack, you weren't using it right - it was lifting from the front enemy trenches too soon and giving the enemy time to mow down your troops. WWII was the same. In the Korean and Vietnam wars, there were actually incidents where a battalion was overrun and the commander would call for an artillery barrage centered on his own command post - his men knew where the foxholes were and so most of them would survive a barrage that killed nearly everything above ground...

Compared to those days, our friendly fire losses are tiny. I think our losses due to accidents are rather low too. Our losses to Iraqi action are amazingly low, and that makes the others look larger. What I would really like to see is a comparison between the total deaths in our deployed forces, and the rate of accidental deaths a military force the same size would typically see in peacetime - from what I remember of the rate of drunk driving accidents when I was in the Air Force long ago, it's possible that this war _directly_ saved American lives...

Anyway, technology is supposed to come to the rescue pretty soon. If plans now underway work out, every American on the battlefield will have his location continually measured by satellite and fed into the computers, and echoed to the whole force, so everyone with a heavy weapon will have this data in his targeting computer. Eventually, even a rifleman may have position markers for his comrades projected on his helmet visor.

Of course, it won't help the Brits much if they don't get compatible systems...

Posted by: markm on April 12, 2003 07:39 PM

Well, J.M., I wasn't trying to impress anyone at all. I was just reporting that every time I've heard a military person refer to what you call "friendly fire", it's as fratricide. Make of that what you will.

Posted by: David Perron on April 14, 2003 10:52 AM

Alas, I don't have the reference handy, but if Gulf War II shapes up like Gulf War I, markm may be right: in 1991, overall loss of life was significantly below what would have been expected if there had been no deployment. The scarcity of alcohol indeed made a huge difference.

I too have been very much impressed by the efficiency of the coalition operations. Those who bemoan our level of losses to friendly fire lack perspective -- historical, operational, and mathematical.

Posted by: Cronaca on April 14, 2003 06:06 PM

Dear Dr. Weevil

congratulations for this thoughtful and enlightening piece of statistics. Really, Thanks, wish I'd have thought of this myself. (Though i prefer the example form medical diagnosis usually given in statistics textbooks). However, since no one else seems to volunteer, i'd like to play the devil's advocate here and point out a few things from your argument, which i feel are not wholly valid (though i concede your claim that the high percentage of deaths due to friendly fire is partly a statistical artefact).
1. "(b) the very high accuracy in targeting" does NOT "produce a shockingly high percentage of deaths from 'friendly fire'." The real cause is "(a) the huge disproportion in casualties" as you can see by simply reducing the accuracy of the coalition troops and noticing that this will make the number of friendly fire deaths rise and not fall as you seem to imply.
2. Your 99.9% success in avoiding friendly fire is problematic, since i a) don't see a way to measure this and b) think it would be pretty pointless even if you could measure it, as it should be 100% when the troops stay at home and thus does not speak of the effectivity or accuracy of your troops (similarly for the case of a siege versus taking a city by force). Rather, the 99.9 percent should refer to the overall accuracy of coaltion forces (you seem to mix these throughout your argument). This is of course also dependent on the sort of combat activities they engage in, but it is measurable and everyone can relate to that statistic (=say what it means), which i assert is not the case with "99.9% success in avoiding friendly fire".
3. From 2, it follows, that the 0.1% of "misses" of coalition forced should include civilian deaths, and even assuming coalition forces are equally likely to fire at their own comrades as they are to fire on civilians, would result in only 5 incidents of friendly-fire, resulting in 25% of coalition losses due to f-f.
4. The latter result is well compatible with results from the second Gulf War, as reported here http://www.counterpunch.org/cook1223.html , here http://www.sci.fi/~fta/stats.htm or here http://www.msnbc.com/news/852294.asp . I couldn't find any statistics for the current war, and googled the others, so these numbers may not be wholly correct. Anyway, they suggest that of 10,000 people killed in the last Gulf War, 9980 were Iraqi and 20 coaltion forces (assuming a ratio of 1:500). Last time the ratio of civilian to soldier deaths among Iraqis was 1 in 3, but let's make this 1 in 10 for this time because coaltion forces seem to be trying really hard to avoid civilian casualties. That would be roughly 9000 Iraqi soldiers and 980 civilians then. Last time, 25% of coaltion forces were killed by f-f, making five in our current example. To me, this means that coaltion accuracy is 90%, and yes, i expect them to do better.
To sum up my argument then, you are not correct in saying the 99.9 accuracy is responsible for the high percentage of deaths due to f-f and you are also wrong claiming that the only way out of the statistical paradox would be to further improve on the 99.9 percent which you consequently declare as next to impossible. In fact, accuracy of coaltion troops is around 90% and i think we'll agree that this can be improved. Still, the kill ratio means that f-f will make up a high percentage of coaltion losses, even if the accuracy is improved, as i conceded in the beginning.
Imho the best way to reduce coaltion losses (and other casualties as well) would be not to wage war unless it becomes absolutely necessary.


Posted by: brmic on April 17, 2003 09:42 AM

Dr. Weevile is right 110%

Posted by: John Smith on October 31, 2003 05:52 PM

There are two ways to ensure zero fratricides occur: 1. Never employ a weapon. 2. Have every weapon you employ engage only enemy forces. If you are, in fact, engaged in hostilities then the first choice will result in unacceptibly high losses due to enemy action - and is therefore not really an option. So when engaged in war, you strive for the second choice. How did we do in the two Iraqi Wars? Let's look...
Before we get specific, we need to understand that fratricide can occur in one of four combat environments: Air-to-Air (AA), Air-to-Ground (AG), Ground-to-Ground (GG)(sometimes called Surface to Surface so as to include maritime assets), and Ground-to-Air (GA). Let's just look at the microcosm of AG frat stats...
In Desert Storm, there were over 2500 aircraft that flew over 126,000 sorties. Of those, 33,462 were dedicated to attacking targets on the ground. Most of those sorties flew against mutiple targets or made multiple passes, but for simplicity let assume one lethal engagement per sortie. Of those sorties, there were 10 incidents of fratricide resulting in 20 deaths and 26 wounded personnel. 10 out of 33,462. That means 99.97% of the time we were employing weapons from the air to the ground away from friendly positions, vehicles and personnel.
What about OIF? There were 29,206 AG sorties flown and 9 fraticide incidents. 99.97% again.
These are not hypothetical numbers. Our troops, our weapons, and our tactics and procedures are that good.
The original premise stated by Dr. Weevil is correct: The reason the Desert Storm stat of 24% is so high is because our total losses in combat was (relatively) so low - 146. There were 466,985 personnel deployed to Desert Storm. We brought home (coincidentaly) 99.97% of them.

Posted by: RV on July 8, 2004 02:49 PM