Rabu, 02 Januari 2013

Data Dive Reveals 15,000 New Victims of Syria War

The world already believed Syria's civil war to be monstrous, with nearly 45,000 slain. But when the United Nations plunged into the disparate databases cataloguing the victims, it discovered there had been an awful oversight. The true death toll was more like 60,000 people, the data-mining operation revealed. And even that elevated total is likely to be low.

The brutal truth is that no one really knows how many Syrians have died in dictator Bashar Assad's brutal crackdown: warzone death estimates are notoriously imprecise. By its own admission, the death toll compiled by the human rights tech group Benetech, on behalf of the U.N., is inaccurate. But its assessment has the virtue of specificity, a factor that preempts some of the doubts raised about mortality estimates in other warzones. The U.N. High Commissioner for Human Rights, Navi Pillay, called the study 'a work in progress, not a final product.'

Benetech's report (.PDF), released on Wednesday, dives into databases of the dead compiled by six Syrian organizations ' several of them tied to the rebellion, such as the Syrian Revolution General Council ' and one by the Assad government. The firm sifted through over 147,000 records to identify and exclude duplicated mortality accounts. Only the 'unique' accounts of 'identifiable victims' fall into Benetech's count. If the records don't show someone's name, date and location of death, Benetech doesn't count it.

That data sift leads to a more precise picture of the Syrian civil war than the rougher estimates previously on offer. Between March 2011 and November 2012, Benetch tallied 59,648 'unique killings.' Over 76 percent of them are male, 7.5 percent are female, and records 16.4 percent of them 'do not indicate the sex of the victim,' Benetech reports. The highest concentration of the dead are found in Homs, rural Damascus, Idlib, Aleppo and Daraa. Benetech is more cautious about estimating patterns of ages for the dead, a stat that casualty-counters often use to make inferences about combatant casualties (on the presumption that 'military-age males' in their 20s through early 40s are likely to fight), since it finds a 'high proportion of missing age data' across the seven datasets. Other U.N. stats also show the war has produced over 478,000 refugees.

Benetech's study suffers from selection bias, as its researchers concede up front. The patterns it finds cannot account for 'unobserved and unidentifiable' killings. If the seven databases do not document a death, or do not document it with sufficient specificity ' say, a name is only partially recorded, or a location of the death is absent ' then Benetech discounts it from the study. Nor does Benetech have a means of fact-checking what's in each dataset, as brutal wars do not lend themselves to such analytic rigor, so an element of faith is unavoidable here. Duplicated information is useful for establishing patterns in the killings, but shouldn't be confused with a verification mechanism for the dead. Accordingly, 'the statistics presented in this report should be considered minimum bounds,' Benetech specifies.

Still, Benetech's emphasis on precision isn't always on display in warzone death counts. Controversially, the British medical journal Lancet sought to calculate the human cost of the Iraq war using a metric called 'excess mortality,' (.pdf) which sought to count dead people who wouldn't otherwise have died in peacetime, whether from American bullets or more inferential measures like poor wartime sanitation. The accordingly high death estimate ' as high as 200,000 dead by 2004 alone ' became a political lightning rod. It's arguable that analytic fears of wading into the debate about how best to measure the death toll has inhibited knowing precisely how many people the Iraq war killed.

Other warzone death estimates are also highly disputable. The U.S. does not disclose how many people have died in the drone attacks in Pakistan, Yemen and elsewhere, prompting observers to guesstimate off of media reports. Various death tallies have come in for criticism for over- or under-modeling the human impact of the wars: some stand accused of undercounting civilians, relying on inaccurate early reports, undue credulity about sourcing and so forth.

Benetech provides almost a clinical agnosticism about its death sourcing. It does not shy away from using the accounts of interested parties ' both the Syrian government and its adversaries ' and explicitly says in the case of Assad's count that it has no reason to disbelieve those figures. (Though it specifies,' we see that the government data source shares very few records in common with the other data sources.') It's single-mindedly interested in death claims that have high degrees of specificity attached to them, a metric that perhaps unexpectedly led Benetech and the U.N. to offer an higher estimate of the Syrian bloodshed than the one typically bandied about in the press.

But the tech firm hints that a 'subsequent report' will wade into precisely the controversies that have marked other warzone death estimates. 'Benetech will address the question of undocumented killings directly through statistical modeling,' it pledges. Get ready for a fight over the validity of that grim model.



Tidak ada komentar:

Posting Komentar