U.N. Reports 35,000 Iraqi Civilians Killed in 2006 Colin Rowat, an economist at the University of Birmingham in England, talks about a report issued Tuesday from the United Nations that estimates nearly 35,000 civilians have died violently in Iraq last year.

U.N. Reports 35,000 Iraqi Civilians Killed in 2006

  • Download
  • <iframe src="https://www.npr.org/player/embed/6872435/6872436" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript

NEAL CONAN, host:

More than 100 people were killed today in Baghdad following a series of bomb blasts in and around the Iraqi capital. The deadliest attack occurred outside of Baghdad University, as teachers and students left classes. Two blasts there killed at least 65 people, with over 100 wounded.

Today's bloodshed comes just hours after the United Nations issued a report that estimates nearly 35,000 civilians died violently in Iraq last year. Joining us now to talk about the report is Colin Rowat. He's an economist at the University of Birmingham in England and studies research methodologies, among other things. He's with us on the line from his office in Birmingham, and it's nice of you to be with us this evening.

Mr. COLIN ROWAT (Economist, University of Birmingham, England): Oh, my pleasure, Neal.

CONAN: Today's report by the U.N. that more than 34,000 civilians were killed last year, nearly three times higher than the number reported by the Iraqi government. Where does this discrepancy come in?

Mr. ROWAT: Oh, very easily. Both of these figures, both the Iraqi government figures and then the U.N. figures, are based on what are called passive surveillance methodologies. And the passive surveillance methodology just waits for officially recorded deaths to reach you.

So in the case of the health ministry or the Iraqi government figures, they're just using fewer sources than the U.N. figures, but it should be stressed that both of these are passive surveillance methodologies, so they will both probably, or almost certainly, in fact, underestimate actual deaths because there'll be more deaths that are not recorded than there will be falsely recorded deaths.

CONAN: So in other words, the U.N. figure much higher than the government figure, and you say they're probably both low.

Mr. ROWAT: Oh, yes. No, I would fully expect that.

CONAN: And what about the issue of gathering this data? It's not easy to find this information…

Mr. ROWAT: Well, that's one of the…

CONAN: …dangerous in others.

Mr. ROWAT: That's quite correct, yes, and that's one of the reasons that these methodologies will almost certainly be undercounts. An alternative methodology is to do a proper population survey. And so in a proper population survey, you collect researchers, and they go out, and they randomly interview people in the population and ask them about mortality that's occurred, or deaths that have occurred in their families, and then you scale up the sample.

So suppose you interview 1 percent of the people in the country, you would then multiply your figures by 100 percent to get an estimate of the overall mortality. So that - those sorts of studies have also been done. They've been much less frequent. There've been two done since the invasion in 2003. They produced much higher figures and, in both cases, though, the researchers conducting those surveys would like to see larger-scale studies of this sort done.

CONAN: And to get even better information, they acknowledge that there's more information out there to be had.

Mr. ROWAT: Oh, that's correct. I mean, I think both of these studies that have been based on proper survey methodologies have been almost heroic. You've had people wandering effectively through a war zone trying to collect data from families on deaths that have occurred in their families. So these researchers' lives are at risk when they do that. So they've given us the best, I think, that they can, given limited resources.

CONAN: We're talking with Colin Rowat, an economist at the University of Birmingham in England who studies research methodologies, and we're talking about the estimates of civilian casualties in Iraq, and you're listening to TALK OF THE NATION, which is coming to you from NPR News.

And that's another word that comes into these estimates, civilian casualties. How do people determine who's a civilian and who isn't?

Mr. ROWAT: Well, I think in the case of the full survey methods, that's not been asked, and one of the reasons for that, obviously, was that asking that question might further put researchers' lives at risk.

So I think otherwise, if you're relying on these passive methodologies, you can just accept what people are telling you. So if the ministry of health reports that this many civilians died, or if hospitals report that this many civilians died, that's what you're stuck with. So you don't have any way of going back to the ultimate source and determining that.

CONAN: So I'm just inquiring about these different methodologies. You have the passive methodologies, which you say the U.N. uses and the Iraqi government uses, but they use different numbers of sources.

Mr. ROWAT: Yes.

CONAN: There's the survey method, which was done - this is basically the public health model?

Mr. ROWAT: Well, that's right. I mean, it's much more general than that. Any politician, for example, will use pollsters, so anyone who's doing any sort of attempt - when you're trying to collect data from a population, you can either ask everyone in the population about what you're looking for - this is incredibly expensive, so even wealthy countries will only conduct full censuses once every 10 years.

So otherwise, if you're trying to get a sense of what's going on in your population, you'll survey a limited sample of the population. So you might survey one percent. You'll want to obviously survey as large a sample as possible, but you'll have time constraints and cost constraints.

So that's the alternative, and this is widely used. It's not just used in public health, but is obviously used in economics, in political science. SO anytime you try to get - or market research. Any of these things will survey a randomly selected population.

CONAN: And then extrapolate from…

Mr. ROWAT: Exactly.

CONAN: Right. Now, there's another group that collects data on civilian deaths in Iraq, the Iraq body count. Which sort of methodology do they use?

Mr. ROWAT: Oh, that's passive. So in their case, they rely on news reports in major English media online. So they have people who are scanning Web sites of major English media outlets, and when they find deaths recorded in two different sources, those deaths will make it into their methodology. So this will be a severe undercount, as well.

CONAN: So that's likely to be a severe undercount, as well. So effectively, you really can't judge the numbers generated by any of these methodologies.

Mr. ROWAT: Well, one of the features of the passive methods - so there's obviously debate about which is the more appropriate technology or way of assessing the numbers.

CONAN: Well, as you know, this is more than just a scientific, academic dispute. This has obviously involved a lot of politics.

Mr. ROWAT: That's right. No, it's highly politicized - but I think there are legitimate questions as well about - so one of the advantages of the passive technologies like the Iraq body count method, is that you get updates fairly quickly. So as news sources are reporting deaths, you can add those to your figures pretty immediate. These may be underestimates, but you're getting information at least flowing in real time, as it were.

CONAN: In days and weeks after the event.

Mr. ROWAT: Sure. As soon as the media report it, as soon as their people find it on a Web site, they'll add it to their figures.

CONAN: Because this is significant, for example, the explosion of the Grand Mosque in Samara is said to have triggered a wave of sectarian violence. This obviously - measuring civilian deaths after that is a way of measuring that.

Mr. ROWAT: That's right. So one of the - so the advantage, of course, is that you're picking up on things quickly versus a survey, which can maybe only be conducted once a year and so will give you very slow feedback. But the disadvantage is that the sensitivity is low. You'll be missing a lot of deaths.

So if you use both techniques in combination, and in fact this has been done by the people who are doing the full-survey technology, they compared trends. So they looked, when they did the survey, at what sort of past pattern of deaths they were finding, and they found that this correlated fairly highly with what people like the Iraq body count were finding.

So it seems that in both cases, they're picking up on the same underlying pattern of deaths, but in the case of the Iraq body count and the other passive methodologies, you're expecting those to be understated, whereas hopefully, with a properly executed survey, they won't be understated.

So it'll take you longer to get the figures, but you'll get a more accurate figure, you hope.

CONAN: And one of the criticisms that comes up with these kinds of studies is they have too wide a confidence interval. What does that phrase mean, and is the criticism valid?

Mr. ROWAT: Well, so the confidence interval - let me just give you a bit of background. Suppose there's a population of 100 people, and you're trying to find out how many of those people ate ice cream last week, if you want. And you can only talk to 10 of them. And so in speaking to the 10, you find that five of them ate ice cream last week. So your best guess is going to be that 50 people in the overall population of 100 ate ice cream last week. You're just going to scale up your sample of 10.

However, you recognize that it's not going to be precisely 50 people. There's some uncertainty around that. And so if you apply appropriate statistics, you might determine that 42 to 58 people with - if you want to be 90 percent confident of 95 percent in your estimate, you might actually estimate between 42 and 58 people last week ate ice cream.

So that 42 to 58 is the confidence interval. Now, the more people you speak to, the narrower that confidence interval becomes. So if you've actually spoken to all 100 people in the population and 47 of them have answered that they ate ice cream last week, then you can be exactly sure that 47, if everyone's telling the truth, ate ice cream last week.

But as you draw up the number of people that you interview, that confidence interval is going to become wider. So you always want to interview more people if you can, to tighten the confidence interval, but if you're constrained, you may not be able to interview a lot of people.

So having a broad confidence interval is simply saying that you don't really know terribly well. You may be doing the best that you can, but given the resources that you have, the best that you have may not be terribly good.

So it's a very honest way of being - it's a way of being honest with your uncertainty. And so the people who have done these proper surveys have made it very clear that they would like to see a properly funded, independent follow-up of their surveys. They recognize that they're operating with limited resources, and such an independent study has not yet taken place.

CONAN: Colin Rowat, thanks very much for your time.

Mr. ROWAT: My pleasure, Neal.

CONAN: Colin Rowat is an economist at the University of Birmingham in England, and he joined us from his office there. I'm Neal Conan. This is TALK OF THE NATION from NPR News in Washington.

Copyright © 2007 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.

NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.