December 20, 2001

MARC HEROLD'S SINS, pt. 2

MARC HEROLD'S SINS, pt. 2 -- THE 108-MAN PARAGRAPH

Here's another randomly chosen example of the kind of "data" that impresses the Guardian. Of Prof. Herold's 3,767 confirmed, "multiple source" fatalities, a significant number -- 108, or 3 per cent -- are footnoted as being drawn from this single, unsourced, unconfirmed paragraph by the Independent's Robert Fisk -- reporting, as he did throughout the war until his recent beating, from a refugee camp in Pakistan -- with no other corroboration, and no apparent effort to rule out multiple counting with other reports. Herold marks them all down in the "Dec. 2" column (Why? Who knows why?):

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...

Oh, yeah. Now that's research.

Posted by BruceR at 09:14 PM

SPEAKING AND REMOVING ALL DOUBT,

SPEAKING AND REMOVING ALL DOUBT, vol. 1 -- MARC HEROLD AND THE AFGHAN CIVILIAN CASUALTIES REPORT

New Hampshire prof Marc Herold's report on civilian casualties in Afghanistan, which claims total Afghan deaths are now higher than the American Sept. 11 casualties, is the talk of the peacenik sites today. On the surface, it looks substantial, although obviously biased, with a complete database of all civilian deaths (in HTML, Word, and Excel!), graphs, etc. to illustrate and back up its statistical claims. However, it's all sizzle, no steak. Once again, a social science prof shames academia by showing no capability for statistical research whatsoever. (As is becoming usual, see Penny for another look at this same story.)

It would take a long time for me to work through the whole of the raw data, but I'll give just one example of Herold's method for now. Taking one day's fatality figure as an example of the entire data set, the report claims 39-42 civilian fatalities in Afghanistan on Oct. 7, the first night of bombing. This is deduced by adding up the results of 5 apparently separate instances, mentioned in 6 separate news reports. I've tried to trace back those reports, and here's what I can say with authority about that total. Two of the news agencies quoted (ITAR-TASS and the Pakistan Observer) do not have online archives going back that far, so I can neither confirm nor deny. Of the remaining 4 sources, two have good archives, but reports could not be found in them on the dates Herold states they were written (The Peshawar Frontier Post and PNS. This could be bad footnoting, or just bad archiving by the sites. If anyone else finds them there, let me know.) That leaves two traceable footnotes an independent observer can evaluate. One, is from the Asheville (Herold spells it Ashville) Global Report News, itself a second-hand compendium. Their actual report is itself unsourced, but is likely drawn from some major news service. The quote is:

Shortly after the US air strikes began on Oct. 7, their home was destroyed in the blast from a bombed arms depot and his two brothers were killed. “We haven’t had war like this before,” said the bare-foot 12 year-old, one of around 2,600 Afghans living in Killi Faizo... Pakistan.

Although there is no definite date information here, and the only info is a single 12 year-old refugee, and there's no idea who the reporter is or a second source of any kind, Herold counts this as two confirmed fatalities in Kandahar on Oct. 7... despite the fact this is probably going to produce double-counting with this and the other, larger, Kandahar death totals he gives on other days.

The other confirmable source for Day 1 is the BBC. Herold attributes them with a claim of 22-25 fatalities in Kabul on Oct. 7. Because his footnote is undated, the exact report he claims supports that number is untraceable, but the BBC does have good archives overall online, and it's fairly easy to find what they DID say about the Oct. 7 bombings at the time:

The Taleban said there were civilian casualties, with about 20 people killed including women, children and elderly people. The Taleban ambassador to Pakistan told Reuters the "horrendous terrorist attacks" had killed at least 20 people across the country. There was no independent confirmation of that figure... Earlier the Taleban said there were about 20 casualties in Kabul, including women, children and elderly people... [Kabul] residents said bombs fell near residential areas, destroying two houses...

That quote, or something close to it, translates in Herold's work into 22-25 confirmed deaths in Kabul that night. Note the conflation of casualties with fatalities, and Kabul with the rest of the country.

Based even on this limited sample, there's no doubt Herold's work is hardly reliable. Of his 39-42 "confirmed" fatalities on Oct. 7, it proved impossible to trace even one to a specific location: certainly the true number was lower, by a factor of 2 or more. A random sampling of footnotes for the other days indicates these shortcomings are shared by the entire dataset. While there was no doubt many civilian casualties in Afghanistan, attempting to replicate just Herold's first-day results seems to indicate the total is only a fraction of what Herold claims... and nowhere close to the WTC fatality list, as if that ever mattered.

However, there is validity in an exercise like this, if it was to be done right. The news agencies that put together such thorough recountings as the WTC story below are behind the 8-ball on coming up with a more reliable civilian casualty figure, which would be informative to everybody. It's in the gap that this leaves that charlatans like Prof. Herold step, giving themselves a 15 minutes in the spotlight their work clearly does not deserve on its merits.

Posted by BruceR at 06:12 PM

WHEN MAINSTREAM JOURNALISM REDEEMS ITSELF,

WHEN MAINSTREAM JOURNALISM REDEEMS ITSELF, vol. 1 -- USA TODAY, 'FOR MANY ON SEPT. 11, SURVIVAL WAS NO ACCIDENT'

A must-read summary of much of the knowable details of the World Trade Center disaster. If you live or work in a high-rise, you want to know the information that's in here, if only to better understand what your best odds are. Elsewhere online, some may have asked about those elevator workers who beetled with everybody else rather than help extract people to the very end, citing the 1993 helicopter-drops of elevator mechanics on the WTC as evidence of more noble elevator-engineer behaviour. My personal take would be that was a ludicrous expectation to be placed on non-uniformed, private sector personnel, of any kind, in 1993 or now. It would surely make far more sense to train special police, fire or even army reserve units as elevator extraction teams, if your city has a lot of high-rise real estate to protect. As the New York emergency personnel demonstrated so manfully in September, only our government should ever be called to regularly put employee lives on the line for the greater good: the best and most responsible thing for most private citizens, regardless of their skills, to do is just minimize the amount of saving that's required for them personally by getting quickly to safety and staying there.

Posted by BruceR at 03:32 PM

HE'S ALSO A CUNNING LINGUIST,

HE'S ALSO A CUNNING LINGUIST, I HEAR

Dawson... Dawson... Dawson... you see sex organs everywhere, don't you? You know, sometimes a medieval astrological observatory is just a medieval astrological observatory. It's not like we're Anarchy Online... now there's a subliminal sex logo.

Posted by BruceR at 02:19 PM