A little forensics: Perusing a recent issue of Frontline, India's premier magazine of the left, one finds a cover story on the war in Afghanistan authored by Professor Marc Herold of the University of New Hampshire. Herold is best known as the author of various studies which compile counts of civilians killed by foreign forces in Afghanistan. His findings have been cited by various newspapers as well as more critical writers such as John Pilger, Ed Herman (who cites Herold's "careful analysis of press reports") and Seumas Milne. Professor Herold also had a chapter in the 2003 Project Censored book and is heavily cited in a book called Strategic Terror by a person with the unlikely name of Beau Grosscup.
It also seems that Herold is the inspiration for the somewhat notorious outfit Iraq Body Count, which like Herold seeks to tally reports of civilian deaths. While the IBC's figures were a badly needed corrective to early reporting on the Iraq invasion and occupation, as time went on their tally - consisting essentially of media reports - was arguably superceded by survey-based research. The difference between their respective findings was great, yet Western media basically ignored the serious studies, instead citing IBC counts as more authoritative. IBC themselves did not disavow anyone of this spurious notion. In fact, IBC even weighed in and publicly questioned the survey-based research despite having no professional credentials in the field.
Herold himself has also come under some fire from various quarters. Media Lens, known for their extremely careful Chomsky-like analysis, took Herold to task after he asserted that the actual number of civilian casualties in Afghanistan could not be more than twice the number reported in the international press. Herold gave no compelling reason for his belief, which would indicate that the ratio of reported to unreported killings is lower in Afghanistan as compared to Iraq, where the proportion of reported versus actual killings is perhaps 1 to 10.
Professor Herold has also had to endure at least one baseless public accusation of being an anti-Semite. In an August 2007 posting, the (then) anonymous blogger Afghanistanica wrote that Herold had attacked Weekly Standard writer Joshua Muravchik for his "Jewish trickery" (Afghanistanica's phrase).
As his blog post had offered no actual evidence to back up the accusation, I contacted Afghanistanica by e-mail, asking for evidence for such a serious charge. When the (anonymous) response offered no evidence, I followed up, again seeking either evidence or a correction. Afghanistanica again offered no evidence, only saying that we should "agree to disagree."
Later, Afghanistanica revealed himself to be one Christian Bleuer, a PhD student, and moved his blog and its archives to a new web address. However, at some point the blog archives were expunged of the Aug 31, 2007 blog entry accusing Professor Herold of racism. Forever lost in the sands of time, it would seem. Except for the Internet Archive, that is, which keeps copies of past web pages long after their owners take them down. So you can see here the original 2007 blog post while here you can see the blog's archives as they have survived, cleansed of Bleuer's smear against Herold.
Returning to Professor Herold's article in Frontline, it is largely boiler plate, offering few surprizes to readers of this blog. It does, however, have one curious passage:
How should one assess Obama’s Afghan war on the basis of the metric of civilian casualties? The U.S. media and the U.S. Left are largely silent (the latter choosing to ignore data I have provided , choosing instead to rely upon questionable accounts by Human Rights Watch and UNAMA)...
[Note 28 reads:]
28. For example, by Dave Markland and Tom Engelhart, see Dave Markland, “Afghanistan Past & Present”, ZNews (June 9, 2009) at [url].
Upon reading that I at first thought, 'Perhaps there is another Dave Markland -- one who lives in the US and who also writes about Afghanistan. Perhaps it's famed volleyball coach Dave Markland of Lenoir-Rhyne University in South Carolina.' Alas, no, Herold is referring to me, merely assuming that I am an American.
Next I thought, 'What was I thinking when I "relied upon questionable accounts by Human Rights Watch and UNAMA?" Surely I would not have been fooled by questionable accounts!' But when I clicked on the link to my article, I found that in fact I quote HRW and UNAMA for purposes quite different from their civilian death tallies per se. I cite HRW for their scathing comments on US military statements while UNAMA is cited for its finding that foreign forces had killed more civilians than the Taliban in the first half of 2007. (Since Herold does not tally insurgent-caused deaths, it is not possible to cite his findings in place of UNAMA's.)*
So, Professor Herold's comment about my article is quite incorrect. Yet there's more. What of Herold's data, reckoned by him to be less questionable than that of HRW, et al? I thought I would check the accuracy of some of Herold's data regarding civilian casualties in Afghanistan - namely, his compilation of reports of civilians killed in US operations as part of the initial invasion of Afghanistan from Oct 7, 2001 to Dec 10, 2001. It is for this data set that he has been most frequently cited.
Herold's data set, entitled "Appendix 4: Daily Casualty Count of Afghan Civilians Killed in U.S Bombing Attacks," attempts to record all media reports of civilians killed by foreign forces during the first two months of Operation Enduring Freedom. Many different incidents are listed, ordered by date, with some detail added in.
In order to check Herold's reliability, I fact-checked all of the entries that I could confirm or deny using Lexis-Nexis. Any entry whose sources could all be checked in this manner were included in the sample data set. This amounted to 22 out of 331 separate fatal incidents, or about 6% of all entries in the Herold data set. Together, the sample data set of 22 record up to 437 dead civilians.
Below, I have reproduced the relevant bits of the Herold data set with my findings from the sources cited added below each incident inside square brackets.
The results of this fact-checking show that Herold's data set is highly unreliable, with his data frequently disagreeing with his sources. Only one of the 22 entries is completely correct, meaning that the information Herold provides is fully confirmed in the sources cited. Four entries entail undercounts which total 127 missed dead civilians while nine entries entail overcounts which total 110, resulting in a net undercount of 17. Thus, the sources Herold cites in fact report up to 454 dead civilians, rather than his assertion of 437.
Annotated sample** of data set of Professor Marc Herold: "Daily Casualty Count of Civilians Killed in U.S. Bombing Attacks." Annotations in square brackets.
-Residential area, Gardez Paktia. 10 dead. WP 10/14/01 [The source cited ('U.S. Bomb Misses Its Target, Hits Residential Area,' Washington Post, Oct 14/01) relates: "In a separate incident, reported yesterday by a source with information from Afghanistan, about 20 Afghan civilians were killed during the bombing of the city of Gardez on Wednesday." Ignoring the question of the quality of this evidence, it is apparent that the entry contains an undercount of 10.]
- Villages of Darunta, Torghar, Farmada Jalalabad area, Nangarhar. 28-100 dead. Indep 10/14/01
[Here's what the source cited ('It was if the rocks themselves were on fire,' Independent, Oct 14/01) actually says: "Something went terribly wrong at the end of the week. In conversations with refugees, a string of names come up again and again: Darunta, Karam, Torghar, Farmada – insignificant villages where, according to consistent accounts by eyewitnesses, as well as those of the Taliban propaganda machine, hundreds of civilians were killed." Herold's range of 28-100 does not come from the cited source; thus the entry undercounts by at least a hundred.]
- Village near Kabul airport. 4+ dead. bomb off course due to technical error AFP 10/24/01 [Consulting the source cited ('List of incidents of US bombs striking non-military targets,' Oct 24/01) we find that the wire report, contrary to Herold, puts the incident (including the mention of technical error) on October 13. Yet on October 13, Herold lists an incident "2 Kms so. of the airport" in Kabul which killed four people. Even more troubling, on October 14, Herold lists an incident in Kabul which "missed airport by 1 mile" and killed four. Overall this results in an overcount of 8+ and a dating error or two.]
- Arghandab town, no.of Kandahar. 10 dead. Indep 10/13/01
[The source cited ('Air Strikes on Afghanistan: Casualties - Witnesses Confirm That Dozens Were Killed In Bombing,' Independent, Oct 13/01) says: "The Taliban reported other civilian deaths but these could not be confirmed among refugees in Peshawar. According to the official Kabul news agency, at least 10 people were killed" in Arghandab. Thus, the number of victims should perhaps read "10+" and the source should really be cited as the Taliban via the Independent.]
- Kunduz. 18 dead. AIP in The Scotsman 10/22/01
[Lexis Nexis did not turn this up. AIP is the Afghan Islamic Press, which began life in 1982 as a mouthpiece for mujahedin. It is commonly said to relay Taliban propaganda. I was able to find an Out There News article which relays the AIP report of 18 dead civilians in Kunduz, but it is clearly the same article referenced two cells above the Kunduz incident, wrongly refering to an incident in Kandahar which is obviously a mislabeling of the Kunduz incident. Thus the data set may have a referencing error and surely overcounts by 18 by doubling the Kunduz incident.]
- Kandahar. 10 dead. father buries son Telegraph 10/15/01
[The source cited ('The planes came at nine . . . at 2am I buried my son,' Telegraph Oct 15/01) does indeed tell of a man burying his son who was killed in an airstrike. However, the article mentions no other civilians killed, thus the entry contains an overcount of nine.]
- Qargha area residences Kabul. 6 dead. 6 killed in house Guardian 10/13/01, Indep 10/13/01 [The Independent ('Witnesses Confirm That Dozens Were Killed In Bombing,' Oct 13/01) cites the Taliban's news agency to the effect that bombs "destroyed homes in Karaga, north of Kabul," but gives no casualty count. The Guardian ('Media war: Taliban propaganda move alarms allies: TV crews are granted visit to bombed village,' Oct 13/01) also cites the Taliban news agency as reporting that "homes were destroyed in Karaga, north of Kabul," but also gives no casualty count. However, Xinhua News of Oct 12 cites the Afghan Islamic Press as reporting that six people died in Qargha "when a bomb hit their houses." Thus the entry should cite AIP (or Xinhua) as its source.]
- w. Afghanistan Bagdhis.  dead. bombing 12 killed and 32 injured AP 10/15/01, citing Taliban data
[I could find no mention of the incident on the Associated Press wire for Oct 15/01. However, Agence-France Presse ('Taliban says US strikes kill 12 civilians but opposition says no deaths,' Oct 15/01) has this: "The head of the Taliban-run Bakhtar News Agency, Abdul Hanan Hemat, earlier told AFP that 12 civilians died and another 32 were injured when a residential area in the capital of Badghis, Qala Nau, was hit overnight Sunday [October 14]." Thus, the entry should read "AFP" rather than "AP" and the incident date is incorrect.]
- Bibi Mahru village, near airport Kabul. 10 dead. bomb kills Gul Ahmad, wife, their chidren + 2 neighbor's children Guardian 12/1/01
[The source cited ('US planes rain death on the innocent: "Precision" raids kill residents in capital city,' Guardian, Dec 1/01) agrees completely with Herold's rendering. This is the only completely accurate entry in the sample data set.]
- Arghandab town, no of Kandahar. 10 dead. Herald Sun 10/19/01
[The source ('US bombs "hit homes,"' Herald Sun Oct 19/01) mentions five dead civilians in Kabul, 12 people killed in Kandahar, then notes: "Planes also targeted Arghandab, a small town about 20km to the north-west, witnesses said." No casualty count is given. Thus the Herold data set overcounts by 10.]
- Tarin Kot Uruzgan. 20-32 dead. Indep 10/25/01, R.L. Barry
[The source cited ('Families Blown Apart, Infants Dying. There Terrible Images of this "Just War" Damage,' Independent Oct 25/01) reports that the incident killed 12 in one household, eight in another, making simply 20 in total. The 20-32 range given by Herold is a misreading, resulting in an overcount of perhaps 12.]
- Tractor and trailer near Tarin Kut Uruzgan. 20+ dead. 20 people inc. 9 children killed when tractor and trailer they were using to flee was hit AFP 10/24/01
[I could find no AFP report to that effect for Oct 24/01, but the Times that day said "up to thirty" had been reported killed by refugees in Pakistan. The Washington Post says "at least 29" people were killed there, again citing refugee claims. Thus Herold's data set appears to undercount by 9 and contain a sourcing error.]
- Macrorayon housing district Kabul. 6 dead. Guardian 12/1/01
[I could find nothing about this incident in the Guardian, Dec 1/01. Thus the data set contains a sourcing error and perhaps an overcount of six.]
- Kandahar area. 8 dead. Indep 11/15/01
[I could find no mention of this incident in the Independent, Oct 15/01. Thus the data set contains a sourcing error and perhaps an overcount of eight.]
- Gluco village, near Khyber Pass Nangarhar. 7+4=11 dead. villagers killed on 19th (7) and 20th (4) Telegraph 11/21/01
[The facts cited are correct, except the date of one bombing. It should read 19th (7) and 18th (4).]
- Kunduz front line Kunduz. 40 dead. hit mud houses AP 11/21/01
[I could not find an Associated Press report on Nov 21/01 that referred to the incident. Several other sources did, including Xinhua: "U.S. warplanes on Wednesday continue to bomb Kandahar and Kunduz, the last two strongholds of the Taliban militia, killing at least two civilians, according to the Afghan Islamic Press (AIP)." Other sources which mentioned the incident reported that AIP said that the number killed was unknown. Thus the data set overcounts by up to 38 and contains a sourcing error.]
- Khanabad area Kunduz. dozens dead. Indep 11/23/01
[I could find no mention of the Khanabad incident in the Independent of Nov 23/01. However, the Afghan Islamic Press reported "dozens" of dead civilians as announced by a Taliban spokesman. See AFP, Nov 23/01, 'Dozens killed and injured in US bombing of Kunduz: AIP.' The date indicated for the incident is November 23. Thus the data set contains a sourcing error and a dating error.]
- Kariz village Kandahar. 6 dead. village destroyed. Hit home of Mahmut, kills his 5 children Indep 12/4/01
[I could not find anything in the Independent Dec 4/01 on the Kariz incident. However, the Mirror ('5 Kids Killed By US Bomb,' Dec 4/01) reports on five dead children (sons of Mohammed Khan) in Kazi Kariz, eight miles south of Kandahar on Nov 28. Thus the data set contains a sourcing error and an overcount of one.]
- Talkhel and Balut villages Nangarhar. 50 dead. villages in White Mtns. Indep 12/2/01, CNN.com 12/1/01
[I could find no mention of the incident in the Independent of Dec 2/01. However, CNN ('U.S. denies bombs hit Afghan villages,' Dec 1/01) does indicate that an airstrike on Nov 30 killed 50 civilians. Thus the data set contains a sourcing error and a dating error.]
- Kili Sarnud Hamlet near Tora Bora Nangarhar. 50 dead. hamlet destroyed Indep 12/4/01 - Fisk
- Lashkargah Kandahar. 46 dead. Indep 12/4/01 - Fisk
- Bibi Mahru. 12 dead. Indep 12/4/01 - Fisk
[Here's everything Fisk ('The River of Victims Runs through Another War,' Independent, Dec 4/01) has to say about civilian casualties, as he reports on refugees arriving in Chaman, Pakistan: "From all over the countryside, there come stories of villages crushed by American bombs; an entire hamlet destroyed by B-52s at Kili Sarnad, 50 dead near Tora Bora, eight civilians killed in cars bombed by US jets on the road to Kandahar, another 46 in Lashkargah, 12 more in Bibi Mahru." It is unclear why Herold has Lashkargah located in Kandahar (when it is the capital of Helmand) nor why the eight dead civilians killed on the road to Kandahar are not included. It also unclear why Herold has grouped Kili Sarnad and Tora Bora together as one event when Fisk seems to refer to two incidents. Thus there appears to be an undercount of eight and some recording errors.]
* For this blog, I have always treated ALL civilian death counts as gross underestimates of actual violence, as they tend to count only those deaths somehow registered or publicized by family. Studies of occupied Iraq show a large difference between recorded victims and actual victims, and (contra Herold) there seems little reason to suppose that deaths are more frequently recorded in Afghanistan than in Iraq.
** While my annotated data set, drawn from Herold's data set, is called a sample data set herein, note that it is not a random sample, though it is of course a systematic one.