From a new paper by Jeffrey Clemens:

In the mid-2000s, U.S. anti-opium policy intensified with a goal of reducing the resources available to Afghan insurgents. To achieve this objective, I show that opium suppression efforts must accurately distinguish between insurgent and non-insurgent suppliers. The required level of accuracy will be particularly high if demand for opium is inelastic and if the insurgents’ initial market share is large. Empirically, I show that demand for Afghan opium is relatively inelastic, that the market share of Taliban-heavy areas is large, and that enforcement has primarily impacted non-Taliban territory. Consequently, anti-opium efforts have significantly increased the drug-trade resources flowing to the Taliban.

I seem to be on a “unintended consequences of social engineering” kick these days. Somewhat ironic that the majority of my field experiments are yielding excellent social engineering results.