Chris Albon over at War & Health pointed me to a very interesting article on a recent scientific study that found the distribution casualties per attack in given years across several different wars followed a power-law distribution. The authors use this observation to create a simple mathematical formula, which Chris has conveniently posted, to predict the number of attacks causing some number of casualties in the next year. The trick in the formula is raising the product of (average number of attacks per year)(number of casualties per attack) to a constant power-law exponent, in this case -2.5.
I write a lot about power-laws in terms of network degree distribution, and I find this observation fascinating, but like any good research the ultimate outcome is more questions. What is the relation between the organizational structures of these conflicts and the distribution of attacks (very hard to answer, given the data requirements)? To what extent can this model be fit to high-intensity conflicts (seems to me they are only fitting it to low-intensity)? Can the model be made richer by adding extremely granular conflict data, e.g. the MNFI SIGACTs database (similar to what Palintir claims to have conducted)?
One worry about such a model, though, is how policy makers can interpret it. The Iraq conflict has been the most statistically analyzed war in our history due to the availability of data; making it easy for policy makers to “benchmark” success or failure. With this model, policy makers could use a predicted number of casualties as a means to assess progress. For example, falling 10 casualties short of the predicted number could be interpreted as the mark of a successful policy; however, this is very problematic and ignores the political reality of low-intensity conflicts like Iraq and Afghanistan.
Automatically Generated Related posts:
Interesting but there are a few things that are not clear to me. Is it the average number of attacks per year or the average number of casualty causing attacks? Casualties or fatalities?
The math baffles me as well. All seems well and good if you want to know how many attacks (assuming an average of 1000/year) will result in ten casualties (around three). However if you use the formula to see how many attacks will cause one casualty you get 1000. Or does that mean that each of the 1000 attacks will result in at least one casualty?
My head is spinning!
[Reply]
I think the model’s simplicity is both a strength and weakness, as you point out. I also agree that you make a very good point about the difference between fatalities and casualties, and my suspicion is that the authors are confusing the definitions.
[Reply]
I’m wondering about the details because I’d like to see if I could apply it to attacks on NGOs in conflict areas. It would be a useful planning tool for places like Afghanistan and Sri Lanka.
[Reply]