Death, Statistics, and Politics

October 13, 2006

Posted by Jay Livingston


It’s hard to wrap your mind around large numbers, especially when they refer to things you’re not familiar with. I remember when my son was learning about dinosaurs in kindergarten (when did dinosaurs became such an important part of the early curriculum anyway, and why?). I couldn’t really grasp the difference between “165 million years ago” and “65 million years ago,” even though a difference of 100 million years is a long time and even though it made quite a difference for the dinosaurs— a difference between being dominant and being extinct.

A couple of days ago, the British journal The Lancet published a study estimating that 600,00 Iraqis had died from violence since the U.S. invasion three and a half years ago. That works out to 470 deaths a day, every day. The confidence interval was broad, so the low-end estimate was “only” 420,000, a number which still sounds incredibly large. (That confidence interval also means that the upper-limit of 790,000 was as likely as the low-end figure.)

Obviously there are political implications in the Iraqi death rate. Arguments in favor of the war would seem a bit weaker if the blessings of liberty which the US invasion brought to Iraq also included far more violent death than Iraqis suffered under Saddam. Here’s President Bush at a press conference the same day.

QUESTION: A group of American and Iraqi health officials today released a report saying that 655,000 Iraqis have died since the Iraq war. That figure is 20 times the figure that you cited in December at 30,000. Do you care to amend or update your figure? And do you consider this a credible report?

BUSH: “No, I don’t consider it a credible report. Neither does General Casey and neither do Iraqi officials. . . .But this report is one -- they put it out before. It was pretty well -- the methodology is pretty well discredited.

QUESTION: So the figure's 30,000, Mr. President? Do you stand by your figure, 30,000?

BUSH: I, you know, I stand by the figure a lot of innocent people have lost their life.

Bush is, of course, wrong. The methodology (“cluster sampling”) has not been discredited. Even so, many people find it hard to imagine the 470-a-day figure. Yes, the news carries reports of violent death every day, but the numbers are smaller. Today’s news reports eleven killed at a satellite TV station. Sometimes the numbers are as high as 50. But 470 every day —is that plausible?

Yes. The news rarely reports on the killings outside of Baghdad, and it rarely reports on isolated killings of smaller numbers of people. Baghdad (about violent 100 deaths a day) is the largest city, and it holds 10% of the Iraqi population. But there are many, many other cities; some of them even make the news reports occasionally — Fallujah, Baquba, Ramadi. Blogger Juan Cole notes that the authorities in Basra admitted last may that people there were being assassinated at the rate of one an hour, 24 a day. And none of those deaths was reported in the US news (or any other Western press).

You can’t use news stories to arrive at statistical estimates. That’s why you need science-based techniques like cluster sampling. The results may at first seem hard to imagine —science always has a hard time when it comes up against “common sense”— but it’s also hard to imagine such a thing as “165 million years ago.” Which may be part of the reason that most Americans don't believe in evolution either.

No comments: