Counterinsurgency spreadsheet corner

Hmm are you sure about this lads - Spencer Ackerman and Matthew Yglesias swap numbers on COIN in Afghanistan, civilian lethality of. I excerpt the numbers bits below ...

What the U.N. has found since 2005 is that civilian casualties have been on the rise as a feckless and underresourced mission came into conflict with a resurgent and adaptive Taliban. The period of January to June 2007 recorded 684 civilian casualties; the figure during that time period rose to 818 in 2008; and 1013 from January to June 2009.

But UNAMA, beginning in its mid-2009 report, noticed that the proportion of responsibility for the civilian casualties was changing:

In the first six months of 2009, 59% of civilians were killed by AGEs [Anti-Government Elements; that is, insurgents] and 30.5% by PGF [Pro-Government Forces; that is the U.S., NATO, Afghan-government forces]. This represents a significant shift from 2007 when PGF were responsible for 41% and AGEs for 46% of civilian deaths.

I bolded the word "significant" there because it's a technical term being used loosely. In fact, 30.5% of 1013 is 309 and 41% of 684 is 280. In other words, coalition forces have been killing about 300 civilians in a six month period at a fairly consistent rate and all the variance in the percentages is in the insurgent violence. In general, don't take percentages of percentages, don't calculate variances of percentages, don't in general do any processing to your raw numbers after you've turned them into percentages, unless you're 100% sure that you know what effect this will have...

So what happened to civilian casualties when McChrystal arrived in June 2009? According to UNAMA’s January 2010 report, its most recent:

Pro-Government forces – Afghan National Security Forces and International Military (IM) forces – were responsible for 596 recorded deaths; this is 25% of the total civilian casualties recorded in 2009. This is a reduction of 28% from the total number of deaths attributed to pro-Government forces in 2008. This decrease reflects measures taken by international military forces to conduct operations in a manner that reduces the risk posed to civilians

596 in a year is as near as dammit 300 in two six month periods. No variance. Nothing changed.

Update, 8:12 a.m., April 16: Derrick Crowe points out in comments below a new USA Today story documenting ISAF backsliding on civilian casualties. ISAF troops “accidentally killed 72 civilians in the first three months of 2010, up from 29 in the same period in 2009, according to figures the International Security Assistance Force gave USA TODAY.” By McChrystal’s own reckoning, then, the system is blinking red and new measures have to be put in place

72 in three months is actually a halving of the run rate; the fact that apparently there were (309-29=) 280 civilian deaths caused by the coalition in Q2 09 after only 29 in Q1 09 shows you that the three-monthly data is just too volatile to give a signal, so I doubt that it will be setting off any red lights.

But but but ...

But the actual problem here is that it's another type 3 error. As I've remarked regularly, "civilian" and "combatant" are not categories known to medical science; no doctor ever said "Congratulations Mrs Sherzai, a big bouncing baby combatant". Somebody categorised these people, after they died, in order to compile the initial number of "civilian" casualties that all these proportions are calculated off.

And it seems to me highly unlikely that the Afghanistani people will necessarily agree with the Army's taxonomic scheme. Recall the discussion of Taliban labour contracts a year ago. There's all sorts of people who are basically farmers, but who from time to time out of poverty, nationalism or intimidation, find themselves planting a bomb or carrying a gun or otherwise involved in an attack (a very large proportion of the output of the Northern Irish media industry used to deal with the tragic dilemmae of this sort of character). When one of them gets shot, then they're definitely a combatant ... or at least I think they are. I am not volunteering, however, for the job of telling the dead guy's family that he was a combatant, because I suspect they'd disagree. If the strategy is about winning haerts and minds, I don't think that the neat distinction between civilians and combatants is actually all that useful.