Apples and Oranges
Posted by Ilan Goldenberg
Not to be obsessed with this issue (but it’s true I am). I did a little more analysis comparing the GAO’s numbers and the Pentagon’s numbers on “sectarian incidents”, which indicates that the Pentagon is likely comparing apples (mid to late 2006 numbers) to oranges (2007 post surge numbers). Again, this is only an indicator. I can’t assert anything for sure, there could be a problem with the GAO numbers. But this goes beyond just the exclusion of certain attacks and may demonstrate that when they started excluding certain types of violence in 2007 they didn’t necessarily do the same for 2006.
Specifically, I took the GAO's daily attacks (page 10 in this testimony) on civilians and converted it to monthly (Note: the GAO's source is the Pentagon). These numbers are a best estimate based on eyeballing and measuring the numbers from the graph so we know they aren't completely exact but close. I compared that to the Pentagon's numbers on "Sectarian Incidents" in the June report. I compared June 2006 to April 2007 (The Pentagon report doesn’t go beyond April). Sectarian incidents are defined by the Pentagon as "threat and violence with apparent sectarian motivations." Two issues:
Abnormality A: Generally speaking civilian attacks reported by the GAO are higher than sectarian incidents reported by the Pentagon. This makes sense since "sectarian incidents" are a subcategory. Except in November and December 2006 sectarian incidents are higher than civilian attacks and in January the numbers are basically equal. I guess that is possible if you include "threats" but it seems highly unlikely. Especially when you consider that for all the months before and after, civilian attacks are significantly higher than sectarian incidents. The end result is that “sectarian incidents” seem extraordinarily high in the three months right before the surge begins.
Abnormality B: As of March 2007, you start to see a steep and pronounced difference between civilian attacks and sectarian incidents, which makes it look like sectarian incidents have suddenly dropped dramatically relative to attacks on civilians. Those type of differences do not exist in the 2006 numbers which track each other much more closely
Anyway, I don’t know exactly what it means, but it seems odd and it has the effect of making the late 06 numbers look really bad and the March 07 and on numbers look much better.