Predicting and Preventing Gun Violence: An Evaluation of READI Chicago

摘要 年轻的黑人男性被枪杀的可能性是白人男性的20倍，更多的黑人男性失去了 死亡人数比排在其后的九大主要死因的总和还要多(CDC，2020年)。芝加哥，它看到了 2016年和2020年的凶杀案激增了50%以上，这是这个问题的一个鲜明例子：90%以上的案件涉及 一支枪支。就像在许多城市一样，受害者绝大多数是来自少数弱势群体的年轻黑人男性 邻里关系。尽管这种暴力行为产生了极高的社会成本，但很少有干预措施 已经考虑了找到最有可能参与枪击事件的人，并向他们提供 提供经济、行为和个人支持服务，而不是更多的执法。 该项目是一项随机对照试验(RCT)，旨在对枪支风险最高的男性进行新的干预。 暴力，旨在测试一种行为知情的、以社会服务为基础的方法。快速就业和快速发展 发展倡议(READI)确定芝加哥男性参与枪击事件的风险最高 通过三种方法：基于行政逮捕和受害记录的机器学习预测； 在所服务的社区工作的街道外展工作人员的转介；以及对离开的人进行筛查 监狱和监狱。然后，READI提供18个月的有支持、有补贴的工作以及认知行为 治疗(CBT)和个人发展规划。READI建立在CBT先前RCT的证据基础上- 基于发现它可以显著减少暴力和其他犯罪行为的编程(Blattman等人， 2017；Heller等人，2017)，这表明有可能减少社会代价最高的暴力形式 而不会招致加大执法力度的附带成本。 在三年多的时间里，2456名男性被随机分配到提供READI的治疗组或对照组 团体免费提供其他可用的服务。这是个人层面枪支暴力事件中最大的RCT 到目前为止进行的干预。这项研究的主要目标是衡量对严重暴力的影响结果 使用行政逮捕和受害记录的涉案和其他犯罪行为。丰富的数据 来源和实验设计也让我们有机会了解每一次招聘有多好 方法预测实际的枪击和暴力参与。由于可能存在一种权衡， 参与者的风险水平和他们对治疗的反应性，我们还将分析风险的变化， 招聘方法的接受率和项目影响教会了我们关于社会最佳目标的知识。 最后，我们将结合影响结果、定性数据收集和效益-成本比较得出结论 关于城市努力解决严重暴力的巨大社会代价的更广泛的政策影响。

Abstract Young Black men are 20 times as likely to be fatally shot than their White counterparts, and more lose their lives to homicide than the next nine leading causes of death combined (CDC 2020). Chicago, which saw homicides spike by over 50% in 2016 and 2020, is a stark example of this problem: over 90% of cases involved a firearm. As in many cities, the victims are overwhelmingly young Black men from a handful of disadvantaged neighborhoods. Despite the extremely high social cost generated by this kind of violence, few interventions have considered the effectiveness of finding those most likely to be involved in shootings, and providing them with economic, behavioral, and personal support services instead of more law enforcement. This project is a randomized controlled trial (RCT) of a new intervention for men at the highest risk of gun violence, designed to test a behaviorally-informed, social service-based approach. The Rapid Employment and Development Initiative (READI) identifies men in Chicago at the highest risk of being involved in a shooting via three methods: machine learning predictions based on administrative arrest and victimization records; referrals from street outreach staff working in the communities served; and screening among those leaving prison and jail. READI then provides 18 months of supported, subsidized work alongside cognitive behavioral therapy (CBT) and personal development programming. READI builds on evidence from prior RCTs of CBT- based programming that find it can dramatically reduce violence and other criminal behavior (Blattman et al., 2017; Heller et al., 2017), suggesting it may be possible to reduce the most socially costly forms of violence without incurring the collateral costs of stepped-up law enforcement efforts. Over three years, 2,456 men were randomly assigned either to a treatment group offered READI or a control group free to pursue other available services. This is the largest RCT of an individual-level gun violence intervention conducted to date. The study’s primary goal is to measure impact results on serious violence involvement and other criminal behavior using administrative arrest and victimization records. The rich data sources and experimental design also gives us the opportunity to learn how well each of the recruitment methods anticipates actual shooting and violence involvement. Since there could be a trade-off between a participant’s risk level and their responsiveness to treatment, we will also analyze what the variation in risk, take-up rates, and program impacts across recruitment methods teach us about socially optimal targeting. Lastly, we will combine impact results, qualitative data collection, and a benefit-cost comparison to draw out broader policy implications about cities’ efforts to address the enormous social costs of serious violence.