Friday, August 24, 2007

When Metrics Don't Show What the Right Wants Them To Show

Kevin Drum does what Kenneth Pollack and Michael O'Hanlon should have done before proclaiming success for "the Surge":

Is the surge working? That is, even if you ignore the lack of political progress, are we even making tactical progress? Since violence in Iraq tends to be seasonal, the only reasonable comparison is one between summer 2006 and summer 2007, so I went to the latest Brookings Iraq Index to check out the most recent numbers.

No figures are available for August, and the surge wasn't completely up and running until June, so the best comparison is between June/July 2006 and June/July 2007. I'm not pretending this is conclusive or anything, but the news sure doesn't look very good. The two tables below tell the story.

The story told is that the surge is working, writes Matthew Yglesias, "... if your definition of "working" doesn't require a decrease in violence or an increase in the viability of Iraq's basic infrastructure."

Matthew continues:
... Of course, even if you saw continued deterioration on those fronts, you might still take solace in good news from the political front except that there . . . isn't any good news on the political front.

Last, I would remind readers that the summer of 2006 was worse than the summer of 2005 which, in turn, was worse than the summer of 2004. Meanwhile, at the time the summer of 2004 was conventionally considered to be very bad situation. We've managed to fail to badly that less-intense forms of failure now look like progress if you squint hard enough.

Given that Kevin's data just comes straight from the Brookings Iraq Index project, one wonders how it is that Brookings fellows like Peter Rodman, Michael O'Hanlon, and Kenneth Pollack seem so unaware of it. Surely the Brookings communications staff should be capable of getting this information into the hands of the organization's own staff.

One of Matthew's readers points out how war supporters manage bad news like this by moving the goalposts:
Whether the way of recording casualties in Iraq or the delivery of electricity, when the data show significant problems growing the data is simply changed.

So, after Linda Bilmes used Veterans Administration data to show physical casualties had passed well beyond 50,000 by January, the Defense Department formally complaine to the dean of Harvard's Kenney School and the data was changed to halve the casualty rate. Similarly the disaster that is electricity delivery in Baghdad was covered by stopping the recording of electricity delivery in Baghdad.

The deception [g]oes on from there.

Even the one bit of good news in these metrics is offset by bad news:
Kevin found some GOOD NEWS(tm) in the Brookings Institute statistics. The number of multiple fatality bombings has declined by 25%. Great news!

Now for the bad news --- the number of people killed by the fewer multiple casualty bombings has increased by 19%. Remember that his time frame excludes the 500+ killed in the attacks on the Yazdi villages. So the bombing campaigns are becoming more efficient at killing people with fewer successful attacks. [...]

A comment at Political Animal predicts: "So, of course the administration will use 'Multiple Fatality Bombings' as the appropriate metric for 'success.' "

No comments: