« No Surrender! | Main | My Commenters Is Smarter Than I: Supply Side Edition »

September 06, 2007

Summer Dyin'

One of the shell games used to justify the surge is to compare the summer months to their immediate predecessors. But in the summer, temperatures reach 120 degrees, and it's too hot to leave the house, much less engage in strenuous, sweaty urban warfare. So the relevant comparison wouldn't be summer against spring, but summer against summer. Over at The Angry bear, Frank de Libero created a graph comparing just that:

Gi Iraqdeaths

GI fatalities are, to be sure, only one metric, and one might expect them to increase during the surge in any case. But the pattern is clear: There's been more violence every month this year than in the corresponding month the year before. This holds true for civilians outside Baghdad, too. So here's what the surge can claim: No political reconciliation, and an increase in violence over the year before. That's one impressively turned corner.

September 6, 2007 | Permalink


Huh. I don't really understand the "projected" numbers for August. I also think that we should be using coalition casualties rather than simply US deaths in these kinds of calculations; coalition casualties are, after all, a function of instability and insurgent activity.

Posted by: Rob | Sep 6, 2007 9:26:55 AM

But Bush sez (in Australia) that we're 'kicking ass' in Iraq.

Posted by: JimPortlandOR | Sep 6, 2007 9:51:45 AM

I don't really understand the "projected" numbers for August.

That was calculated by multiplying the 65 deaths for Aug 2006 by the average ratio, Jan-Jul, of 2007 divided by 2006.

Posted by: Frank de Libero | Sep 6, 2007 11:20:23 AM

To be fair, you should break out KIA fatalities from non-combat fatalities. We should expect the number of non-combat deaths to rise with more troops in Iraq and a more forward-leaning posture. A quick cut at icasualty.org numbers gives us the following for hostile and non-hostile kills, 2006-2007:

JAN +36 -14
FEB +27 nc
MAR +45 +4
APR +37 -2
MAY +56 -4
JUN +41 +4
JUL +33 +9
AUG +1 +21 (provisional)

Still doesn't look good for the home team: at best, August rates remain substantially unchanged from last year for hostile deaths, and up significantly for nonhostile deaths.

Now, if we look at the hypothesis underlying the month-to-month data analysis, it can be rephrased as saying that the late summer months (July and August) should exhibit decreased variability in casualty rates, since the climate for those months overwhelms political and military factors. (Obviously, casualty rates are a poor proxy for data such as total numbers of contacts, but it will be quite some time before the more meaningful data is made available to historians.)

Briefly playing around with the data (excluding the invasion months of March, April and May 2003), we get:

MONTH s2 REL_s2 s
APR 859.4 4.559 29.315
JUN 520.33 2.760 22.810
OCT 497.35 2.638 22.301
NOV 485.35 2.575 22.030
MAY 472 2.504 21.725
DEC 446.95 2.371 21.141
SEP 379.35 2.012 19.476
FEB 375.75 1.993 19.384
JAN 258.15 1.369 16.067
AUG 254 1.347 15.937 (provisional)
MAR 250.95 1.331 15.841
JUL 188.46 1.000 13.728

April numbers are skewed somewhat by high April 2004 (Vigilant Resolve) casualty numbers, as are November numbers for substantially the same reason. Removing April 2004 from the numbers doesn't significantly change the high variability of that data, but removing November 2004 makes November the least variable of the months:

MONTH s2 REL_s2 s
JUN 520.33 4.901 22.810
OCT 497.35 4.684 22.301
APR 472.16 4.447 21.729
MAY 472 4.445 21.725
DEC 446.95 4.209 21.141
SEP 379.35 3.573 19.476
FEB 375.75 3.539 19.384
JAN 258.15 2.431 16.067
AUG 254 2.392 15.937 (provisional)
MAR 250.95 2.363 15.841
JUL 188.46 1.775 13.728
NOV 106.16 1.000 10.303

A better way of handling increased Fallujah casualties is to remove those specific casualties rather than the entire month, but I'll leave that to folks seeking more rigor than found in a blog-post comment.

So what does this tell us? Well, if our hypothesized months (July and August) show low variability, then that indicates support for the climatic theory. (July is very low-volatility, August less so but still on the low end.) Second, changes in low-volatility months may indicate that the values of dynamic factors (political, military, etc.) have changed radically enough to counter the stable climate factors that create a "ceiling" on enemy contacts. In this sense, the major change in July 2007 is bad news.

On the other hand, changes in high-variability months will have to be scrutinized carefully to see if they represent significant progress or regress. My personal rule of thumb (after the past few minutes of eyeballing these numbers) is that a change of greater than plus/minus sixteen hostile deaths in September and fifteen in October should be considered potentially significant, with nineteen and twenty-two being significant. (This is almost certainly oversimplified and wrong, but at least it's a place to begin arguing from.)

Incidentally, even if we take the less-optimistic values (19/22), should casualty rates decline by that much, we'll still be where we were in 2005. But regaining stability in Iraq should be seen as an improvement over the worsening course we're currently on.

This is just a very quick, nonscientific and nonrigorous look at the numbers, begging for a more in-depth study. For right now, anyone care to take a whack at my arguments or tear apart my numbers?

Posted by: WatchfulBabbler | Sep 6, 2007 1:47:16 PM

George W. Bush is a genius.

By allowing global warming to continue uninterrupted, he is making a world a more peaceful place!

Posted by: The Confidence Man | Sep 6, 2007 3:11:41 PM

The comments to this entry are closed.