Friday, August 31, 2007

Who to trust?

In an interview with the Australian newspaper, Gen. Petreus is making the following claims on success:

General Petraeus told The Australian during a face-to-face interview at his Baghdad headquarters there had been a 75 per cent reduction in religious and ethnic killings since last year, a doubling in the seizure of insurgents' weapons caches between January and August, a rise in the number of al-Qa'ida "kills and captures" and a fall in the number of coalition deaths from roadside bombings. ....

According to General Petraeus's figures, which will be put to Congress, the number of ethnic- and religious-related deaths would be down to a quarter of what they were last December by the end of August. He said "ethno-sectarian deaths" were the most important measure of progress.

"If you look at Baghdad, which is hugely important because it is the centre of everything in Iraq, you can see the density plot on ethno-sectarian deaths," he said.

"It's a bit macabre but some areas were literally on fire with hundreds of bodies every week and a total of 2100 in the month of December '06, Iraq-wide.

"It is still much too high but we think in August in Baghdad it will be as little as one quarter of what it was."

This sounds great, but can we trust what he is saying as it is massively conflicting with numerous other reporting. For instance, the Washington Post's report on the GAO had, for the GAO at least, a very strong statement that in evaluator speak is a scream that someone is LYING/withholding data (which to an evaluator is almost the same thing)

the draft suggests that future administration assessments "would be more useful" if they backed up their judgments with more details and "provided data on broader measures of violence from all relevant U.S. agencies."

The National Security Network is also taking a look at the previous use of metrics and evaluation criteria and is seeing a very inconsistent pattern on the usage of data.

A number of U.S. agencies differ with the Administration's assessment that sectarian violence is down and in fact there are inconsistencies within the Pentagon's own reporting. The Iraq Study Group concluded that in the past car-bombs that don't kill Americans, murders, and inter-ethnic violence were not tracked in order to demonstrate reduced violence. Recent analysis indicates that some of these trends continue. More importantly, the military has refused to show the public any evidence to support the claim that violence is down.... (emphasis mine)

The original data for the five months before the surge began (September 2006 through January 2007) indicated approximately 5,500 sectarian killings. In the revised data in the June 2007 report, those numbers had been adjusted to roughly 7,400 killings – a 25% increase. These discrepancies have the impact of making the sectarian violence appear significantly worse during the fall and winter of 2006 before the President’s “surge” began.

U.S. data on civilian violence is unreliable and excludes most acts affecting sectarian and ethnic cleansing. " No Iraqi data are reliable and US and Coalition data focus far too much on major bombings, major incidents, killed to the exclusion of wounded, and violent acts to the exclusion of most acts affecting sectarian and ethnic cleansing. No data can be fully trusted in terms of accuracy. More importantly, many current metrics are useful largely as measures for counterinsurgency in a nation filled with diverse civil conflicts and where the most violent insurgent acts are only an uncertain indicator of the trends in security and stability." [CSIS, 8/22/07 ]
The data does not include Shi’a on Shi’a violence in the South or Sunni on Sunni violence in the Sunni Triangle. “The data on the drops in attacks are complex, and it must be stressed that they do not count clashes or violence at lower levels between the tribes and Al Qa’ida or some forms of intra-Sunni Islamist feuding and fighting… These figures also ignore growing Shi’a instability in the south, and particularly in the southeast, and a growing threat from Iran.” [CSIS, 8/6/07 ]

Next the AP and McClatchy have been keeping indepedent tracking systems of fatalities in Iraq. The AP is reporting that by their counts, in the first eight months of this year, more people have died in Iraq than during the entire year of 2006. McClatchy is reporting that they can not verify or reconcile US claims about security improvements in Baghdad with their own information. And the US government is not releasing their methodology.

Furthermore, if we use for total coalition fatalies, we see that in the first seven months of this year, every month's total fatality level was above 2006 levels, and five of the seven months had war time highs in total fatalities. Adding in incomplete August information, August 2007 was deadlier for Coalition forces than August 2006, and it was not the worst August of the war (2005 was by 2 more fatalities).

I work and have been trained as a program evaluator and the Daniel Davies 1 Minute MBA quote is a good heuristic to keep in mind:

  • Good ideas do not need lots of lies told about them in order to gain public acceptance.

  • Fibbers' forecasts are worthless.

  • The Vital Importance of Audit

  • Follow these three rules, and you'll be amazed at how many good decisions you can make with good information, but I digress. Gen. Petreaus is making a set of claims of success that is contradicted by numerous external sources, the GAO is screaming for more data and data transparency, and the few less disputed claims (AQI influence waning) and potentially the number of IED deaths going down are spurius to the strategic objectives of the surge.

    No comments: