A Glaring Error in the Washington Post

Readers of my blog know that I do not hesitate to bring you bad news about casualties in Iraq when the news actually is bad. But when the news is good, I will not hesitate to tell you about that either.

At Powerline, Paul Mirengoff linked to this new Washington Post article on casualties in Iraq:

Body Count In Baghdad Up in June

By Joshua Partlow
Washington Post Foreign Service
Thursday, July 5, 2007; Page A01

BAGHDAD, July 4 -- Nearly five months into a security strategy that involves thousands of additional U.S. and Iraqi troops patrolling Baghdad, the number of unidentified bodies found on the streets of the capital was 41 percent higher in June than in January, according to unofficial Health Ministry statistics.

During the month of June, 453 unidentified corpses, some bound, blindfolded, and bearing signs of torture, were found in Baghdad, according to morgue data provided by a Health Ministry official who spoke on condition of anonymity because he was not authorized to release the information.

The number of unidentified bodies found in Baghdad was significantly higher in June than in January, the month before the deployment of additional U.S. and Iraqi forces began in the capital. However, the overall number of civilians who died in violence in most Iraqi provinces has generally declined since January.

In January, 321 corpses were discovered in the capital, a total that fell steadily until April but then rose sharply over the last two months, the statistics show.

Did the Washington Post forget about their own story from April that reported this information about body counts from January?

U.S. commanders say sectarian murders fell from 1,200 in Baghdad in January to fewer than 400 in March. Markets are reopening, and a few thousand families have trickled back to areas they had fled.

But they agreed that among the most troubling trends in Iraq has been the proliferation of suicide bomb attacks, because they risk reigniting sectarian revenge killings and undermining the government. Suicide bombings have increased 30 percent over the six weeks that ended in early April, according to military data.

As it happens, I think that the number of sectarian murders supplied by the US military for January (1200) might be too high, as I discussed in detail here back in April. But the number in the new story (321) is absurdly low, so the Washington Post needs to print a retraction. In fact, according to the most reliable numbers available, the body count in Baghdad is down by 27% in June compared to January. As such, the headline of the story needs to be changed from Body Count In Baghdad Up in June to Body Count In Baghdad Down in June. I covered those statistics in detail here in the context of a larger discussion about declining civilian casualties in Iraq, but today I'll focus only on the daily body count in Baghdad.

Iraq Coalition Casualty Count keeps track of casualties in Iraq based on deaths reported in the media (you can find their reports here). Every single day, they register the body count for Baghdad. They may not catch every murder, but their methodology is consistent and is therefore able to capture genuine trends. Here, for example, is the article that formed the basis of their report for 1/10/07:

BAGHDAD - Police recovered the bodies of 60 people with gunshot wounds and signs of torture from various parts of Baghdad in the 24 hours to Wednesday evening, an Interior Ministry source said.

There is a report like that for virtually every day in January and every day in June. It is a much more objective way to track trends than relying on anonymous officials who do not explain the basis for the numbers they provide to over-eager reporters. When you add up the numbers reported by the police, you find that the number of dead bodies found on the streets of Baghdad was 735 in January and 540 in June (a 27% drop). This is much like the overall drop in casualties throughout Iraq for the month of June. Here is a graph showing the daily body counts provided by the police and reported in the media:


The Washington Post instead reports a 41% increase between January and June. This is simply an error, and it needs to corrected (see my earlier complaint about how this paper tracks casualties here). The same article also reports these additional Baghdad casualty figures:

But even before the plan went into effect, the number of bodies discovered had fallen well below the levels of last fall. In October, for instance, 1,782 unidentified bodies were found in Baghdad, according to the United Nations, citing official statistics provided by the Health Ministry.

By January, the total dropped to 321 in the capital, according to the statistics provided to The Washington Post, followed by 294 in February, 272 in March and 182 in April. But the figure spiked upward to 433 in May and 453 last month. A Health Ministry spokesman could not be reached for comment on the statistics despite several attempts.

These numbers, which are all over the place, are simply ridiculous. The Washington Post reporters clearly have no sophistication in these matters. The mistake they make -- and it is a huge mistake to make if you are attempting to measure a trend -- is to compare figures that are based on different methodologies. The figure of 1782 for October is based on an unreliable morgue report supplied to the UN by some anonymous official (morgue figures are untrustworthy, they have been seriously challenged, and they are no longer routinely provided, except by unauthorized, anonymous officials). The other casualty figures appear to be provided by different anonymous officials (not by the UN) using an unknown methodology. That is not a sensible way to track trends. To track trends, you need to adopt a single methodology, validate the numbers it provides (my most recent effort along those lines can be found here), and then stick to it. Contrast that approach with the one used by those reporters for the Washington Post. They just grab numbers wherever they can find them, cobble them together, and then imagine themselves to be tracking trends. It isn't sensible to do that.

To get a feel for how inaccurate their numbers are, look at that figure of 182 for April. I don't know where that came from (because the reporters do not say), but it is ridiculously low. For example, here is a story in USA Today that relies on the same UN figure for October (which they should not have done) but comes up with a much different figure for April, again by relying on someone who was not authorized to speak to the media:

The number of unidentified bodies found in Baghdad — an indicator of sectarian violence between Sunni and Shiite Muslims — dropped from a high of 1,782 in October to 411 in April, according to an Interior Ministry official who declined to be named because he is not authorized to speak to the media.

So which is it, 182 or 411 for April? According to Iraq Coalition casualty Count, 425 bodies were found in Baghdad in April. These are based on reports provided by officials who are authorized to speak to the media and who do so every single day. Unless you think the media reports that document these deaths are lies, the number supplied by the Washington Post (182) is way off base.

Here is the New York Times weighing in on the subject:

In Baghdad, 730 civilians were reported killed in June from assassinations, bombs or small-arms fire. That was down from 1,070 in May, a decline of almost 32 percent, an Interior Ministry official told The New York Times.

However, the number of dead bodies found in Baghdad, a measure of sectarian killings, while lower in June than in May, was still higher than in April, according to the Interior Ministry official. In April, there were 411 dead bodies found in Baghdad; in May, there were 726; in June, the number dropped to 540.

The numbers in this article for April, May and June are quite close to the numbers you get from Iraq Coalition Casualty Count (ICCC), which are 425, 649, and 540, respectively. I suspect that these are official statistics of the Iraq Interior Ministry (because the source is not identified as being unauthorized), but I can't be sure. In any case, let's look at the numbers for the month of April from 4 different sources:

Washington Post: 182 ??
USA Today: 425
NY Times: 411
ICCC: 425

Does one number stand out as being different from the others? The new Washington Post story is way off the mark. That being the case, you simply cannot trust the overall analysis provided by that story, and you certainly cannot trust the headline: Body Count In Baghdad Up in June. This headline has it exactly backwards, and it really needs to be corrected.

The Washington Post story also said this:

But the number of unidentified bodies found on the streets is considered a key indicator of the malignancy of sectarian strife. While the declining number of bombing victims suggests that efforts to control violence are showing some success, the daily slayings of individuals, in aggregate, speak to an enduring level of aggression.

As it happens, this "key indicator of the malignancy of sectarian strife" dropped by 27% in June compared to January. How many people are reading the page 1 story in the Washington Post that got it exactly backwards and how many are going to get the rest of the story that I provide here? I shudder to think, so I will do my best to get the Post to correct their mistake. If they do, the correction will probably appear in some isolated paragraph on page A17, and it will simply say that it is hard to get reliable casualty figures from Iraq because various sources of information yield different conclusions. That's what a correction is likely to say in the unlikely event that a correction is made. The truth is that reliable figures can be obtained (read my detailed explanation here), and I have gone to great lengths over the years to demonstrate that this is true. I doubt that any reporter at the Washington Post has put in anywhere near the effort that I have on this issue. I am quite confident that sectarian murders in Baghdad declined significantly in June relative to January. At a minimum, I am certain that the method I use is vastly more reliable than the method used by the Washington Post in their page 1 story.