Bayesian Inferences for Counterterrorism Policy: A Retrospective Case Study of the U.S. War in Afghanistan

Abstract

ABSTRACT This study employs hierarchical Bayesian analysis of terrorist attacks to provide a retrospective analysis of the war in Afghanistan between 2002 and 2018. We examine the relationship between U.S. troop levels, target type, and the severity of attacks in terms of the number killed or wounded. We find that although some targets might have become better fortified after enduring attacks (such as police departments), terrorists would subsequently succeed in either attacking other targets (such as educational institutions) or even those same targets in subsequent years. Our analysis also finds that increases in U.S. troop levels throughout much of the conflict did not seem to quell violence, although explanations of this phenomenon are considerably more nuanced after taking expert opinion into account. We hope our analysis provides a useful retrospective analysis of the U.S. War in Afghanistan for policymakers to assess how or if security measures in the country could have been improved over the course of the conflict. Our modeling framework, however, is easily generalizable to other conflicts worldwide and thus provides a useful statistical tool for analyzing terrorism in many other settings as well.