The theory of public goods is mainly about the difficulty in paying for them. Our question here is this: Why might public goods not be provided, even if funding is available? We use the Afghan Army as our case study. We explore this issue using a simple model of a public good that can be provided through collective action and peer pressure, by modeling the self-organization of a group (the Afghan Army) as a mechanism design problem. We consider two kinds of transfer subsidies from an external entity such as the U.S. government. One is a Pigouvian subsidy that simply pays the salaries, rewarding individuals who provide effort. The second is an output/resource multiplier (the provision of military equipment, tactical skill training, and so forth) that amplifies the effort provided through collective action. We show that the introduction of a Pigouvian subsidy can result in less effort being provided than in the absence of a subsidy. By contrast, an output/resource multiplier subsidy, which is useful only if collective action is taken, necessarily increases output via an increase in effort. Our conclusion is that the United States provided the wrong kind of subsidy, which may have been among the reasons why the Afghan Army did not fight. (JEL A1, D7, D9)