655.000 morts ? Comment ont-ils calculé cela ?

L’étude d’un groupe de statisticiens de la John Hopkins University sur les pertes en Irak depuis mars 2003 (655.000 morts) est l’objet d’une intense polémique. Elle ajoute un élément intensément tragique au discrédit qui caractérise aujourd’hui la guerre en Irak.

Amy Goodman, de la station de radio Democracy Now !, a reçu hier l’un des auteurs de cette étude, Les Roberts. Parmi les questions posées, celles qui concernent la méthodologie employée (avec l’intervention d’un autre intervieweur, Juan Gonzales).

AMY GOODMAN : It’s good to have you with us. Why don't you lay out exactly what you found?

LES ROBERTS: Sure, we, as you said, went to about 50 neighborhoods spread around Iraq that were picked at random, and each time we went, we knocked on 40 doors and asked people, “Who lived here on the first of January, 2002?” and “Who lived here today?” And we asked, “Had anyone been born or died in between?” And on those occasions, when people said someone die, we said, “Well, how did they die?” And we sort of wrote down the details: when, how old they were, what was the cause of death. And when it was violence, we asked, “Well, who did the killing? How exactly did it happen? What kind of weapon was used?” And at the end of the interview, when no one knew this was coming, we asked most of the time for a death certificate. And 92% of the time, people walked back into their houses and could produce a death certificate. So we are quite sure people didn’t make this up.

And our conclusion was comparing the death rate for that 14 months before the invasion, with the 40 months after, that the death rate is now about four times higher. And, in fact, it’s twice as high as when we last spoke two years ago and when we did our first study. So, things have gotten bad, as you stated. We think about 650,000 extra people have died because of this invasion, and about 600,000, some 90%, are from violence.

JUAN GONZALEZ: Well, I’m sure you have heard by now the responses of President Bush and military leaders about this. What is your response to their saying that this is not credible?

LES ROBERTS: You know, I don't want to sort of stoop to that level and start saying general slurs, but I just want to say that what we did, this cluster survey approach, is the standard way of measuring mortality in very poor countries where the government isn’t very functional or in times of war. And when UNICEF goes out and measures mortality in any developing country, this is what they do. When the U.S. government went at the end of the war in Kosovo or went at the end of the war in Afghanistan and the U.S. government measured the death rate, this is how they did it. And most ironically, the U.S. government has been spending millions of dollars per year, through something called the Smart Initiative, to train NGOs and UN workers to do cluster surveys to measure mortality in times of wars and disasters.

So, I think we used a very standard method. I think our results are couched appropriately in the relative imprecision of [inaudible]. It could conceivably be as few as 400,000 deaths. So we’re upfront about that. We don’t know the exact number. We just know the range, and we’re very, very confident about both the method and the results.

(…)

JUAN GONZALEZ: Les Roberts, I would like to ask you something about the methodology of the study. Clearly in Iraq, as in most wars of this type, the level of violence is uneven across the country. It might not necessarily even correspond to the population densities of different areas. What was the methodology that you used to select the particular clusters that you chose?

LES ROBERTS: Sure. That’s a great question. And you’re right. In Iraq, there is a huge difference in death rates between, for example, the Kurdish north, which is relatively safe, and the Sunni Triangle, where the death rates are extremely high. And what we did was we got a population estimate of every government, from the Iraqi government, and we randomly allocated these 50 clusters that we were to go visit proportional to the population in each of those governments, so that, if in the Kurdish north there is only 20% of the population living in the couple safest provinces, we would naturally end up with a sample that’s 20% or so from that zone.

And then, once we had picked that we were going to visit two or three neighborhoods in a certain governance or province, we would then make a list of all the villages and towns and cities, and again randomly pick one of those to visit, so that big places had a larger chance of being visited than smaller places. And then, finally, when we got down to the village level or to the section of a city, we would pick a house at random, visit it and the other 39 houses closest to it to grab a cluster of 40 houses. And luckily, in the analysis, we can sort of look at how much variation there was between clusters.

And when we reported this, we didn’t say it was 655,000 deaths. We said it was 655,000 deaths, and we’re 95% sure it’s between about 400,000 and 950,000. And that range of imprecision is capturing that variance between neighborhoods that you described, some places having a lot of violence, and some not. So there is less than a 2 percent chance that the number is well below 400,000. So, you know, it’s not precise. It’s incredibly hard to do this kind of work in times of war, and I think that this is awfully good, given the conditions.

Mis en ligne le 13 octobre 2006 à 08H49