Thursday, November 08, 2007

Fact Checking the Fact Checker on Iraqi Civilian Deaths

First I should note that it is generally agreed (I think almost universally) that the Iraqi civilian violent death rate declines sharply in September and October 2007. I will however, comment on fact checking of Testimony by General Petreaus and a statement by Hillary Clinton on September 11 2007. One might argue that, as a practical matter, that debate is now academic as we have two more months of information which make things clear. However, that misses two of the important points of fact checking. It is useful to tell who is honest so lies which are no longer of practical importance are worth discovering. It is useful to punish dishonesty so denouncing old lies is useful.

I consider it a clear rule of fact checking claims made on September 11 2007, that people are not labled dishonest for failing to predict the future.

I was very displeased with Michael Dobbs' first fact check of claims about Iraqi civilian deaths (although I messed up the link).
In this article, Dobbs hinted that Petraeus's claims were questionable, but said he would report on their accuracy later. In contrast, he denounced the Betray-Us advertisement.

His second effort
was a very thorough job marred by an unwillingness to draw a firm conclusion. Dobbs said he couldn't tell if Petreaus cooked the books, because MNF-I would not allow him to audit them. I would say the standard is that until they are willing and able to explain the discrepencies between their numbers and independent numbers, they should be considered to have admitted that they cooked the books. If accountability can be blocked by secrecy, we might as well give up on the concept of facts.

His third effort finally reaches a conclusion giving Petraeus and Clinton One Pinocchio each. Thus the result of all the effort is criticism of Petraeus and of two separate critics of Petraeus. I think each was added for balance. The cases against both MoveOn and Clinton are feeble and no other senator or advocacy group was criticized.

I learned something important from the last article. My criticism of the MNF-I numbers is that they are not possible given the huge death toll of the August 14th bombing aimed at Yazidis. I still think this is true of one of the figures presented by Petraeus on September 11, in which the killings were inexcusably classified a non sectarian. However, "according to [the chief military statistician in Baghdad, Colonel David R.] LaRivee, the August figures included several "high casualty" incidents, such as devastating suicide attacks on Yazidi villages in northern Iraq. By U.S. estimates, 326 people were killed in the Aug. 15 bombings of the Yazidi villages. (Preliminary estimates from the Red Crescent spoke of more than five hundred dead.) Had it not been for these mass suicide attacks, the decline in civilian casualties would have been even more pronounced. Most western news organizations based in Baghdad reported a fall in civilian deaths between August and September."

This helps to explain the gap between MNF-I and independent estimates of deaths in August. Iraq Body Count and the Associated Press were probably using the initial higher estimate of deaths on August 14th. The justification for MNF-I's lower estimate of deaths caused by the Qahtaniya bombings is important.

In general, while respected the work Dobbs did, I remain extremely unsatisfied by his analysis. The rules appear to be that Petreaus is to be criticized only if his subordinates confirm that he was dishonest. This is an absurd approach to fact checking. In contrast, Clinton is criticized for alleged vagueness and an aid's refusal to answer an irrelevant question and is pilloried for accurately quoting news reports.

I am going to try to organize the rest of this post. First I will discuss data sets often relying on Dobbs's discussion. Then I will discuss the fact checking of Petraeus, Clinton, The Council on Foreign Relations and MoveOn.

First the data. All of Petreaus's claims are related in some way to estimates of Iraqi civilian deaths made by his subordinates and MNF-I. The overall series shows a sharp decline in deaths from December 2006 (chosen as a peak and being two months before the beginning of the surge) and August 2006. The Associated Press and Iraq Body Count have very different estimates showing a smaller decline from December 2006 till August 2007 and no clear trend from Spring 07 until August 07.

Oddly MNF-I figures show more civilian deaths than casualties in December '06 and January '07 implying a negative number of non fatal injuries. Both strange features of the MNF-I data can be explained by MNF-I's decision to include "host nation reports" of deaths. They therefore count deaths and casualties differently, but did not explain this in the titles and legends of figures. Almost all of the decline in Iraqi civilian deaths is explained by a 70% decline of Iraqi reported deaths. As I noted earlier in this blog, this huge decline occurred after the Iraqis ceased to provide raw data to the UN. Essentially, General Petreaus's claim is as solid as our confidence that ethnic cleansers accurately report how much ethnic cleansing they have been doing.

Excluding host nation reports gives a series very similar to the IBC and AP series. Deaths in August 2007 were slighly more than deaths in February 2007 and there was no clear trend during the surge.

Dobbs made an important contribution by insisting that MNF-I provide the separate series (necessary to explain how they counted a negative number of non fatal injuries and why their numbers are totally different from the AP and IBC numbers). I'd say that claiming to believe such implausible numbers is clearly dishonest. Dobbs does not note as others have that estimates of deaths in a given time period (say December 2006) change later.

Dobbs does not note the Associated Press time series. I assume this is because he doesn't read Talking Points Memo, which, unlike the AP constructed a series. The series shows an increase from the beginning of the surge through August 2007, although basically it shows no trend during the surge.

Dobbs notes that Iraq Body Count shows much less of a decline than MNF-I. I think he should conclude that MNF-I includes extremely dubious Iraqi numbers in it's count and therefore gives a totally different picture than independent sources which attempt to be comprehensive (last qualifier to exclude the Washington Post count of bodies delivered to morgues). Thus their numbers are highly suspect and Petreaus's summary dismissal of all other numbers and refusal to discuss the cause of the difference was dishonest.

Basically there is no way that Petreaus would have convinced anyone (but wingnuts of course) that we should trust the Iraqi government more than the Associated Press and yet is exactly what he did. I'd say it's worth 2 or 3 pinocchios right there.

Dobbs, however, only considers criticisms of Petreaus from his own subordinates. This is absurd but not completely uninteresting since they are not willing to stand by his claims. Just cutting and pasting.

"Civilian deaths of all categories, less natural causes, have also declined considerably, by over 45 percent Iraq-wide, since the height of the sectarian violence in December."
--General David Petraeus, congressional testimony, September 11, 2007.

Asked about Petraeus's "over 45 percent claim," LaRivee said that he would not "want anybody to read too much into the precision of that statement. [Petraeus] knows very well that the host nation reports have some problems with them. But the trends are moving in the right direction."

Petraeus stated the claim as a definite precise fact. Not "according to coalition and host nation reports collected to date" nor with any other qualifier. If he knew very well that the numbers he presented were not accurate to two figures, he is guilty of perjury. The only claim that LaRivee is willing to make some time between October 1 and October 19 is that that has been a decline. This could refer to the decline from August to September, that is, after Petreaus's testimony or from December 2006 till February 2007 when the surge started.


According to Colonel LaRivee, the Iraqi authorities are improving their data collection processes all the time, but the system is still far from perfect. There is "definitely a lag in host nation reporting." Some events that were reported as having taken place in December 2006 could well have taken place in November. There are sometimes duplicate reports of the same incident, leading to inflated figures.

Get that, there is definitely a lag in host nation reporting, therefore, the host nation reported figures for August 2007 were not complete when the power point presentation presented on September 11 2007 was prepared. Therefore, compared to earlier months, the figure for August 2007 was biased down according to Colonel LaRivee. This means that Petreaus's claim was definitely false. He did not qualify the claim with "according to the data we currently have" as would have been required to take credit for the apparent good news due to the bias without technically committing perjury.

That puts his up to 4 Pinocchios. His own subbordinate says that his sworn claim was not true.

For balance and on the general Clinton rules and possibly as a result of Mau Mauing, Dobbs also scrutinizes Senator Clinton's claim that

"The reports that you provide to us really require the willing suspension of disbelief. ... If you look at all the evidence that's been presented, overall civilian deaths have risen."
--Sen. Hillary Clinton, responding to Petraeus during his Senate testimony.

Now I don't see why Clinton alone among people who don't crunch the numbers should be scrutinized, but I also don't see what is wrong with her claim. The statement is not totally explicit as it doesn't specify the time interval over which civilian deaths have allegedly risen. Dobbs interprets it as being true only if it is true for all time intervals, and, in particular establishes as the standard reference point December 2006, which is the month with the highest reported civilian deaths according to MNF-I. This is requiring Clinton to cherry pick against herself. The natural interval is the surge, and, as I have noted a few times above, all sources except for the Iraqi government and the clearly incomplete morgue count support her claim.

Since she was challenging the MNF-I figure it is absurd to suggest that her assertion is false if it is inconsistent with the MNF-I figure. I think a large part of Dobbs problem is that he doesn't know about the ap time series. AP, IBC and MNF-I without "host country reports" are very clear and Clinton is very clearly right unless you think that "declined" must mean "declined since December 2006" or hell if those numbers are refuted how about during the Mongol sack of Baghdad. The death rate sure has declined since then (guessed at about 100,000 a day).

Here is the verdict on Clinton

Hillary Clinton Claimed that Iraqi civilian deaths "have risen" without specifying her terms. She read talking points almost verbatim from a Democratic think tank, the Center for American Progress, which argues that the violence has not gone down. Her spokesman, Philippe Reines, told us earlier that Clinton was referring only to the months of July and August 2007 when she claimed that the death toll had "risen." He declined to respond to a question about whether the senator now believes that Iraqi civilian deaths have risen or fallen since last December.

OK let's see. What basis is there for Dobbs's decision that "decline" must mean "decline since December 2006" ? There is none and there can be none. Clinton's choice of time period is much more reasonable that Petreaus's who obviously chose the very best time period for his argument with no possible justification (wins one Pinnocchio for that).

Worse yet, Dobbs asked his question in October and thus presumes that to be honest Clinton must be pre-cognitive. He did *not* ask Reines if deaths had declined by August 2007 which is the only question relevant to fact checking a claim made on September 11. For all I know, Reines refused to answer for exactly that reason and would otherwise have argued "yes but not during the surge, which is the longest time interval which is relevant to teh debate, according to sources other than Petreaus and his subordinates "

I think Clinton clearly meant to refer to recent changes and never accepted comparisons to December 2006. I think her claim, which clearly referred to sources other than MNF-I is absolutely true. I think she deserves a star and Dobbs 2 Pinocchios for trying to distort what she said and for dragging her in to balance the criticism of Petraeus.

Dobbs criticises the Council of Foreign relations for keeping a claim based on out of date data on its web site after the revision was pointed out to them. I'd say failure to note that the source of the data now makes a different claim invalidating one's assertion is worth a solid 4 Pinocchios. Dobbs gives them one for no reason.

Thus his harshest judgement by far is directed at And why?

He complains that they include a polemical claim in an ad largely devoted to fact checking

"Most importantly, General Petraeus will not admit what everyone knows; Iraq is mired in an unwinnable religious civil war."

"Every independent report on the ground situation in Iraq shows that the surge strategy has failed."

"General Petraeus is a military man constantly at war with the facts."

They are an advocacy organization not fact checkers. I think also that, given the stated aim of the surge strategy, political reconciliation, it clearly has failed. One can only argue otherwise by redefining success.

Also Dobbs denounces as follows

"That's because, according to the New York Times,"

criticism with no reference to The New York Times

"The Washington Post reported that assassinations only count if you're shot in the back of the head -- not the front."

True, the Post has reported this, [STOP right there Dobbs]

Dobbs confirms that the claim made by is absolutely totally true. He blames them for a false claim made by his employer which they accurately quoted.
Clearly he has no sense of Irony. The Washington Post condemns Washington Post. Have people who work at the post no sense of shame ?

"According to news reports, there have been more civilian deaths and more American "
soldier deaths in the past three months than in any other summer we've been there."
True on U.S. military deaths.
an August 25 Associated Press report that war-related deaths in Iraq have nearly doubled over the last year, from an average daily toll of 2006 of 33 to 62 so far this year."

The fact check notes that a claim based on News reports is inconsistent with MNF-I reports, which is sort of MovOn's point.

But this is my favorite

"We'll hear of neighborhoods where violence has decreased. But we won't hear that those neighborhoods have been ethnically cleansed.
Reporters based in Baghdad have reported that one reason for the declining violence is the ethnic cleansing of neighborhoods. See this useful New York Times graphic, for example. Petraeus did not dispute a New York Times report that 35,000 Iraqis have left their homes in Baghdad as part of ethnic cleansing.

Yep your read that right, Dobbs claims that he can fact check a prediction (note the future tense). MoveOn was referring to Petreaus's future testimony. and you know what ? They were right. While earlier MNF-I officers had shown maps of Baghdad color coded to show the ethnic cleansing, Petreaus showed mapps where the change in the number of attacks was illustrated but the color codes all reffered to December 2006 thus hiding information on ethnic cleansing that he definitely had. gets a crystal ball for accurately reporting the future.

Dobbs has no respect for facts, denouncing an organization for claims which he confirms are absolutely true or which are advocacy not verifiable fact while he mildly chides people who make up claims based on nothing, whose subordinates say they know that their sworn testimony was false.

A shame because he works hard. I think it is clear that there is a connection between the good reporting on the MNF-I numbers and the horrible analysis. If Dobbs had not bent over backwards to be unfair to Petreaus's critics and more than fair to Petreaus he would never have gotten the figures on host country reported vs coalition reported deaths.

I think that's the choice these days, people committed to call them as I see them fact checking can't get at the facts.

No comments: