Scholars of civil wars have long argued that non-state actors can use selective punishment to reduce collaboration with state adversaries. However, there is little systematic evidence confirming this claim, nor investigation into the mechanisms at play. In this paper, we provide such evidence from the drone war in Pakistan. Militants in Pakistan’s tribal areas engaged in a brutal counterespionage campaign with the aim of reducing collaboration with the United States. Our analysis combines a novel dataset of collaborator killings with data on drone strike outcomes. We find that strikes killed half as many militant leaders and fighters following collaborator killings and that this suppressive effect likely works by deterring spying in the future. Beyond providing an empirical confirmation of the selective punishment hypothesis, our paper suggests an unacknowledged vulnerability of the drone program to reprisals against local allies and collaborators that limits its effectiveness as a long-term tool of counterterrorism.

内战学者长期以来一直认为，非国家行为者可以使用选择性惩罚来减少与国家对手的合作。然而，几乎没有系统证据证实这一说法，也没有对起作用的机制进行调查。在本文中，我们从巴基斯坦的无人机战争中提供了这样的证据。巴基斯坦部落地区的武装分子进行了一场残酷的反间谍活动，目的是减少与美国的合作。我们的分析将合作者杀戮的新数据集与无人机袭击结果数据相结合。我们发现，在合作者被杀后，罢工杀死了一半的激进领导人和战士，这种抑制作用可能会通过阻止未来的间谍活动发挥作用。除了提供选择性惩罚假设的实证证实之外，