I'm sure everyone has heard about the study done by researchers at Johns Hopkins, estimating Iraqi civilian deaths since 2003 at approximately 655,000. Impressive number, all right - but what does it mean?
Well, the folks over at Iraq Body Count have taken exception to the techniques employed by Johns Hopkins. I find this interesting in particular because Iraq Body Count has been fairly critical of the casualty numbers reported by the Pentagon since the war began, charging that the government has consistently underestimated the number of civilians killed (and there is certainly room for debate on this subject - a lot of which hinges on exactly who is a "civilian?" If the insurgent drops his RPG before I shoot him, does he become a civilian again? Hmmm.)
But I digress. If you read the press release from IBQ, they lay out clearly their concerns. Specifically, Johns Hopkins used a very small sampling of the population of Iraq in order to generate their findings - 47 clusters of 40 homes, for a total of approximately 1800 households. From the responses they gathered, they then extrapolated those results to cover all of Iraq. The problem, as noted by IBQ, is that the violence in Iraq is not uniform throughout the country - some areas (such as those surveyed for this report) have had heavy fighting. Other regions have seen little or no fighting, so extrapolating results from one province to the entire country does not give you anything near an accurate picture. Doing that gives you a result that requires approximately 150 unreported attacks every day for the last three years! Surely someone would have noticed?
To put it in perspective, if I were to visit 47 locations in New Orleans, and interview 40 families at each location and ask which families had lost homes in the last three years, I'm sure that a large percentage of them would answer in the affirmative, especially if I focused my survey on the Ninth Ward, for example. Let's then extrapolate our numbers, and we discover that over 200 million people in America are homeless!
This simply points out why I am leery of statistics. By choosing our sample population, the questions we ask, and the sample size, it is possible to prove just about anything.