August 09, 2002



Just a recap on Herold's new estimate. Readers may remember we tried to trace how much truth lay behind Herold's estimate of 39-42 confirmed fatalities in the first 24 hours of the Afghan offensive... we could not find evidence, outside of a self-serving Taliban press statement, for even one of them. In his latest estimates, Herold has brought his numbers for that night to between 21 and 36. He still counts as confirmed 2 fatalities a story by an unnamed 12 year-old refugee who doesn't even mention which day or where two of his family members died, and still counts 15-20 fatalities in Kabul, based on that same Taliban statement, that 20 people died in the whole country the first day... I suppose that's an improvement.

How many really died the first day? Based on Carl Conetta's statistical sampling that showed Taliban statements overstated on average by a factor of 4 or more what really happened, and given that they claimed 20 fatalities in all of Afghanistan the first night, the number is almost certainly in the single digits. Herold disagreed with other more serious estimates before by a factor of three... I would estimate his new figures are at best still off, overall, by a factor of two. Getting closer...

Posted by BruceR at 11:32 PM



The utterly discredited Marc Herold claims once again that he has the straight dope on fatalities in Afghanistan, in the Guardian... his figures are ludicrous, as always, and deserve no further consideration. What's interesting now is the character assassination he practices on Carl Conetta of the PDA, not to mention Bill Arkin and Human Rights Watch.

Comparison with the PDA... reports is difficult to make as they do not reveal raw data and exactly which sources were employed.

Other than the 70 footnotes in the online appendix comprising his dataset, no, I suppose Conetta doesn't make any mention of exactly which sources...

Conetta's total was lower than mine only because it relied exclusively on western sources.

In fact, Conetta relied on largely European sources, as he considered those the only impartial ones (the 70 reports above). He also specifically criticized Herold for including second-and third hand reports and instead limited his counting of fatalities to reporter eyewitness accounts and family reports of fatalities. Sound statistical sampling, Conetta reported, showed that reports by refugees of non-immediate family fatalities, as well as Taliban claims, were inflating death totals three, four or more times over what turned out to be the reality... about the difference between his claims and Herold's, he politely refrained from noting at the time. Herold does not even try to criticize Conetta's methodology, or his devastating criticisms of Herold's own "study".

Herold out-and-out lies about HRW, too:

HRW officials, it was widely reported, had "said privately" that they estimated the civilian death toll at between 100 and 350 in December...

The quote in question again, from the Globe and Mail's Murray Campbell, noted by this blog back in January:

Human Rights Watch, a U.S.-based organization, offers a rough estimate of at least 1,000 civilian deaths...

The other number, originally cited in the WSJ, actually was made relatively early in the war (the story it's taken from says it was an estimate made "early in December") but by the end of that month, with more information becoming available after the Taliban fled, HRW was already revising their estimates upwards, for a fatality rate closely comparable to those of the other responsible researchers at the time... Herold still prefers to put forward their older smaller estimate here, though, pretending it's HRW's full count in order to make HRW look stupid and discountable.

Herold also lies about the reasons for deflating his own numbers now by over 30 per cent:

A weakness of the initial study was some double-counting due to confused site names - the figure for the October to December period should have been between 2,650 and 2,970 civilian deaths

You can't find in online anymore, but I happened to save his old database from December. To take one example of many, in his initial estimate he claimed the bombing of the Sultanpur Mosque on Oct 10 had killed an estimated 70 people. In the current online database, that is downgraded to 15, with the 70 now given in square brackets... square brackets meaning, we are told, that the number has been "deemed unrelieble [sic] because either [sic] questionable sources or military casualties." I guess it wasn't all double-counting after all (there was certainly no confusion about the site name here)... Herold just apparently doesn't want to admit he initially accepted some wildly inflated figures uncritically.

The guy simply can't tell the truth, it seems. I wouldn't trust him to make change for a chocolate bar, at this point, let alone tell me something important about the wider world.

Posted by BruceR at 05:55 PM