Looking at insurgencies, and data selection

Via Zenpundit, I came across a good article in Armed Forces Journal by Capt. Robert Chamberlain that looks at the faulty idea that insurgencies/counterinsurgencies always take a long time:

It is simply untrue that counterinsurgencies are inherently long-term struggles.

When I started studying the Tuareg insurgencies in Mali and Niger, I was surprised to learn that there had been an insurgency in 1963-1964 in Mali. I had never heard of it, because it hadn't killed very many people. One Tuareg clan - the Kel Adagh, based near Kidal, Mali - rebelled against the Malian government and were brutally crushed in short order. As I wrote in my thesis:

The Malian Army crushed the Kel Adagh rebellion by using widespread violence against civilians, torture, declaring the countryside outside of a few towns a vast free-fire zone, poisoning wells necessary for any activity outside towns, killing Tuareg herds of cattle and camels to force them into the cities, executing the religious and political leaders of both rebel and neutral clans, and threatening napalm bombing from fighter-bombers.

In Chamberlain's metaphor, this was a "kitchen fire" rather than a five-alarm blaze, and because it did not reach 1000 battle deaths, the 1963 rebellion does not appear in the main intra-state war database, Correlates of War (COW).

When academics study insurgencies en masse, they create databases full of very bloody, long insurgencies that kill lots of people but ignore the many many more small, short-lived insurgencies. Consequently our understanding of insurgency as a mode of conflict is skewed towards large bloody messes. I'd hazard a guess that this problem doesn't have the same degree of impact in studying conventional wars, because conventional wars are easier to find in the historical record, and usually the combatants are states that wield significant combat power, meaning there aren't as many kitchen fires that don't turn into five-alarm blazes.

The problem of data selection is cemented in the Correlates of War database. This database is a one-stop shop for researchers on war, but it only counts conflicts that result in more than one thousand battle deaths. Some sort of filter like that is necessary or else the data set would be too large to be useful. The problem comes when researchers like Paul Collier take the COW data, look at intra-state conflict and extrapolate their findings to all intra-state conflict ever, instead of just intra-state conflict with more than 1000 battle deaths.

Using COW data, Collier and Hoeffler reached the conclusion that rebellions happen when there exist opportunities for rebellion (rather than grievances for rebellion). Had they thought about the limitations of their data, they would have realized that their argument only showed that rebellions survive to create conflicts large enough to appear on COW's radar when the opportunity exists for large rebellions - and that rebellions driven by grievances might occur in hostile environments for rebellion but would be wiped out quickly, like the 1963 rebellion by the Kel Adagh.

Dan Byman, the head of my old program, said something related in a talk a couple months ago. He noted that a lot of people argue that terrorism doesn't kill large numbers of people. Byman argued that's only true because when terrorism does start to kill lots of people, we don't call it terrorism any more. Exhibit A, the firebombings of Germany and Tokyo. Exhibit B, the 1990s civil war in Algeria. Exhibit C, the situation in Iraq.

This whole problem is similar to Taleb's idea of "silent evidence" - in order to understand big house fires, you have to look at the big fires themselves, as well as small fires that don't become big fires. Only then can you really understand what factors are important in turning small fires into big ones (click the link to see that even Harvard professors do not understand this concept).

The point of Chamberlain's article was that the mantra that 'counterinsurgency campaigns take a long time' has been used as an excuse for the failure to provide a coherent strategy for a favorable outcome in Iraq and Afghanistan:

According to conventional wisdom, we will be in Iraq and Afghanistan for at least 10 years because they are insurgencies, and that’s how long insurgencies take...
This is an abdication of the responsibilities of strategic leadership. The American public is owed more of an explanation than, “Well, these things take a while.” It is owed a comprehensive strategic vision...

Fabius Maximus continues this point by arguing that when an insurgency has carried on long enough and is damaging enough for a country to beg for large-scale American military involvement, that country is probably doomed to lose anyway. A more effective counterinsurgency strategy for the United States would focus on the small "kitchen fires", and try to keep those kitchen fires from becoming big fires, rather than reacting late to put out big fires.