Fuzzy Numbers... Again

September 18, 2007

Fuzzy Numbers... Again
Posted by Ilan Goldenberg

OK, you would have thought that after all the heat that the Pentagon took for its inconsistent violence numbers they would have been very careful about the latest data that they issued in their September report to Congress. But they weren’t. There are significant inconsistencies between the numbers General Petraeus showed to Congress regarding civilian casualties and the numbers in the Pentagon’s latest reports. Again, Petraeus’s numbers seem to make the period before the surge look worse and the numbers after the surge look much better.

I graphed the Pentgon’s data on civilian casualties from its latest report (Pg 20). The data is for daily casualties, which includes both wounded and killed (I only included Iraqi civilians, which is the green bar). I converted the numbers to monthly casualties and compared it to the civilian death numbers on Page 3 of General Petraeus’s presentation to Congress. I don’t know if the data comes from the same source. Although, to be clear the Pentagon report data comes from the database that Petraeus himself stated was the best source of information and was verified by two U.S. intel agencies. Petraeus’s data comes from “Coalition and Host Nation Reporting.” I graphed everything based on eyeballing the data so the numbers are not exact.

Still, you’d expect the Iraqi civilian death numbers to track the Iraqi dead and wounded numbers and obviously civilian dead should be lower than wounded and dead combined. Unfortunately that is simply not the case. Three observations:

According to the MNC-I data there has been no improvement since either December (The numbers Petraeus and the Administration often cite) or February (when the surge actually began). Why wasn’t Congress shown these numbers in the presentation by General Petraeus? Why only the good news numbers? Why the lack of clarity on Petraeus’s sourcing? Especially since he himself acknowledged that the best numbers come from the MNC-I database.

In terms of actual anomalies

Anomaly A: Somehow in December, the month that is always cited by the Pentagon and the Administration, Petraeus’s Iraqi dead is actually greater than the MNC-I Iraqi Dead + Wounded. That makes absolutely no sense. You can’t have more dead than dead and wounded combined.

Anomaly B: In the months after the surge begins Petraeus’s Iraqi dead numbers are significantly lower than the dead + wounded numbers in the Pentagon report. This is inconsistent with the entire history of the previous year, where the numbers track closely. The only explanation would be a dramatic increase in the wounded to dead ratio. Perhaps there were more car bombings that injured people but didn’t kill them, as opposed to close range executions where victims do not survive. Or maybe there is another explanation. Still it seems inconsistent to see this major split just as the surge begins..

I’ve also graphed what the wounded numbers should look like based on this data (i.e. you can take the total civilian casualty numbers and subtract civilian deaths to find wounded). Notice how in December 2006 and January 2007 the numbers are actually negative, which makes no sense at all. By far the three highest months of wounded are March 2007, April 2007 and August 2007 – after the surge has started.

Overall, the numbers used by Petraeus have the same effect as all the other inconsistencies. They make the numbers right before the surge look extremely bad and the numbers during the surge look much better. Maybe that’s just a coincidence. But it does raise more questions.

September 18, 2007 at 11:25 AM | Permalink

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d83451c04d69e200e54ee07aff8833

Listed below are links to weblogs that reference Fuzzy Numbers... Again:

The comments to this entry are closed.

Disclaimer

The opinions voiced on Democracy Arsenal are those of the individual authors and do not represent the views of any other organization or institution with which any author may be affiliated.
Read Terms of Use

September 18, 2007