The Surge's Untrustworthy Numbers

The big factoid in favor of the surge is that violence is down. As Bush put it a few months back, "Within Baghdad, our military reports that despite an upward trend in May, sectarian murders in the capital are significantly down from what they were in January."

Devil, meet details. The Pentagon classifies violence as "sectarian killings," not simple murders. So those numbers don't count, among other things, Shia on Shia violence in the South, Sunni on Sunni violence -- including between al Qaeda in Mesopotamia and Iraqis Sunnis -- in the North, carbombings, and much else. But even within this vastly constrained definition, the Pentagon, without letting anyone in on their methodology, is changing the definition from month to month. Ilan Goldberg did the lord's work by graphing the Pentagon's numbers from the last few reports, and watch how the numbers for the very same months change with each successive report:

And we're not just seeing random fluctuations -- they're mainly changing downward, in order to reflect lower sectarian violence. But why would the January 2006 be lower in the June report than in the March report? Were the dead resurrected?

"But wait!" You say (because you're rude, and you interrupt a lot). "In the June report, killings were revised upward! That's true. But the timing matters. As Goldberg explains, "The impact here is that it makes the “pre surge” situation look extraordinarily dire and therefore signals progress thereafter."

The shell game here has to do with the term "sectarian murders," which the Pentagon is apparently defining differently from month to month, albeit without telling anyone what's changed. In other words, you can't trust these numbers. But they're the ones that are being used -- and will be used -- to argue for the Surge's success.