by Javier Osorio and Christopher Sullivan
Download the Full Article: What the Afghan War Diary Really Tells Us About the Changing Dynamics of the Conflict
The major news headlines that followed the release of the Afghan War Diary, a collection of tens of thousands of leaked field memos from Afghanistan, describe how U.S. forces are battling an increasingly resilient and well-armed Taliban army. Coverage in the New York Times featured two major stories -- an intensive review of events at Combat Outpost Keating, where U.S. forces defended themselves against an ever-growing and more violent Taliban enemy, and the analysis of a half dozen significant incidents highlighting the changing dynamics of the conflict. Coverage in the Guardian, meanwhile, focused on 200 "key events" that documented significant increases in the Taliban's fighting capacity. Indeed, from these reports it appears as though U.S. efforts in Afghanistan are, at best, maintaining an uncomfortable status quo. At worst, the war is being lost to Taliban forces, whom we now know have been receiving aid from Pakistan.
But how can we draw inferences about a war involving more than 100,000 American soldiers and nearly 10 years of combat based solely on a smattering of documents hand selected by reporters? By purposefully choosing to report just a few hundred documents released as part of the War Diary, and ignoring the broad swaths of data contained in the rest of the records, the existing reportage has opened itself up to charges that the coverage is biased towards the perspective given off by those hand selected documents. A better strategy for understanding what the War Diary can tell us about how the war is faring would be to analyze all of the records and let the data speak for itself.
A statistical analysis of the more than 76,000 events so far released by Wikileaks reveals that the war is not faring as reported on in the major newspapers. The picture revealed is actually much worse.
Download the Full Article: What the Afghan War Diary Really Tells Us About the Changing Dynamics of the Conflict
Christopher Sullivan is a PhD candidate in the Department of Political Science at the University of Notre Dame, where he studies Comparative Politics and International Relations. Mr. Sullivan is also an International Dissertation Research Fellow at the Social Science Research Council, and serves as Managing Editor at The Journal of Conflict Resolution. During the 2010-2011 academic year, he will be conducting field research for his dissertation on state surveillance in Guatemala.
Javier Osorio is a PhD student in the Department of Political Science at the University of Notre Dame where he studies Comparative Politics in Latin America and American Politics. Mr. Osorio's dissertation research focuses on the spatio-temporal dynamics of organized crime Violence in Mexico. He currently serves as Statistical Consultant for the Department of Political Science at Notre Dame.
About the Author(s)
Comments
Mr. Peeples, one need only compare DoD output for public and governmental consumption with the RAW data in these Weakileaks incident reports. One immediately wonders if, as in Vietnam, command is just plain hearing too many "yes sirs" to be able to admit that it's no longer worth the killing of our mom&dad soldiers in trying to create a command learning cure. The first time you try to do an appendectomy with the patient belly down and butt up, one could patiently await a learning cure. But a repeat a second time suggest that you're dealing with a bad SURGE-on doing bad SURGE-ry
I qualify my comments by first pointing out that I have absolutely no qualifications to comment in the first place. That being said...What are the metrics that General Petreus is using to define success in the Afganistan theater? It seems to me that the authors are suggesting that the Taliban are engaging more frequently and that this somehow suggests consequently that "the war is being lost to the Taliban". Could not other alternative explainations be plausible:
1. Is the recent "surge" in troop numbers allowing for American forces to patrol further into hostile areas and subsequently encourage Taliban attacks?
2. Are attacks becoming more frequent because small units are moving into remote village locations to "live among the people" and provide security to rural villages as a primary mission?
3. Are we defining success as the number of engagements our troops find themselves in or in the degree of confidence that the Afgani people have in our ability to control the environment?
I just have a vague feeling that confounding variables might be at work here. I would be very interested in hearing some "ground truth" from those who are taking care of business over there. Is this thing "winnable"? Finally, I'd like to take this opportunity to thank those of you who are taking care of business over there.
The authors are absolutely right to highlight the value of using statistical techniques on datasets such as this one to illuminate trends and relationships, rather than relying on non-random "cherry-picking" of particular reports.
That being said, however, it seems a little unfair to suggest that the inferences drawn by the press were based solely on the items case-studied in press reports of the Wikileaks data. UN and ISAF trend data on attacks and casualties is widely available, and regularly seen by most reporters who regularly cover Afghan issues. Since the SIGACTS information essentially parallels what is already available OS, it is hardly surprising that reporters chose not to focus their reporting on this (unsurprising) aspect of the Wikileaks data, but instead on what seemed rather more interesting vignettes.
Second, while the lagged effects of coalition activity that the authors report are interesting, it might pay to be a little more careful in some of the analytical language being used. How do we know that coalition activities were "deterring" Taliban attacks, as opposed to disrupting them, or degrading Taliban capabilities to launch such attacks? The two are very different things, and indeed the difference cuts to the core of some of the current COIN debates about kinetic actions.
Similarly, I'm not sure that later (2008-) lagged correlations between coalition activities and subsequent increases in Taliban attacks leads one to the conclusion that the former make the "Taliban more violent, not less violent". There are a number of other explanations--ranging from the types of operation, battle rhythms, improved use of ISR to identify potential targets--which would also create a similar apparent statistical relationship. So too would changes in SIGACTS data collection and recording protocols.
None of this is to disagree with the authors' conclusions about the war going poorly, something that has been widely reported in the public data (including the publicly available data on security trends). However, in their very worthy desire to show the value of quantitative analysis of a dataset like this, it might be wise to be a little more careful in the interpretations that they are reading into apparent correlations.