Policymakers and advocates make contradictory claims about the effects of providing military equipment to local law enforcement, but this intervention is not well understood because of severe data limitations and inferential challenges. I use 3.8 million archived inventory records to estimate the magnitude of sources of bias in existing studies of the 1033 Program. I show that most variation in militarization comes from previously unobserved sources, which implies that studies that show crime-reduction benefits are unreliable. I then leverage recent policy changes to evaluate the effect of military equipment: the Obama Administration recalled property under Executive Order 13688, which resulted in a forced demilitarization of several hundred departments. Difference-in-difference estimates of agencies that retained similar equipment show negligible or undetectable impacts on violent crime or officer safety. These findings do not suggest that similar scale federal reforms designed to demilitarize police would have the downside risks proposed by proponents of military transfers.

政策制定者和倡导者对向当地执法部门提供军事装备的影响提出了相互矛盾的主张，但由于严重的数据限制和推论挑战，这种干预并没有得到很好的理解。我使用 380 万份存档的库存记录来估计 1033 计划现有研究中偏倚来源的程度。我表明，军事化的大多数变化都来自以前未观察到的来源，这意味着显示减少犯罪收益的研究是不可靠的。然后我利用最近的政策变化来评估军事装备的影响：奥巴马政府根据 13688 号行政命令召回了财产，这导致数百个部门被迫非军事化。保留类似设备的机构的差异估计表明对暴力犯罪或警官安全的影响可以忽略不计或无法察觉。这些调查结果并不表明旨在使警察非军事化的类似规模的联邦改革会具有军事转移支持者提出的下行风险。